A “FAQ” About Sharky Laguana and Silent September

Sharky Laguana is still advocating against Spotify’s royalty model. His latest post on Medium is called Streaming Music is Ripping You Off. As you can guess from the title, it’s not an exercise in subtlety. According to Laguana, streaming music services like Spotify are saying “Your choices don’t count, and you don’t matter,” and that “You Are Worthless.”

Just as I stated in Sharky Laguana and “Fair” Spotify Royalties, I believe (along with others like David Touve) that Laguana’s numbers don’t add up. He also consistently uses loaded terminology (“click” vs. “listen”), false dichotomies (“10,000 fans who stream a song once, or one fan who streams it 10,001 times”), and other rhetorical mistakes. It is not an article that is making an argument; it is an article that is meant to be inflammatory.

In it, he attempts to generate outrage for a campaign called “Silent September.” This is the plan: “This September, when you aren’t listening to music, put your favorite indie artists on repeat, and turn the sound down low.” The intent is to “make the problem so visible that the major labels feel it would be better to switch to a different system where this kind of manipulation isn’t possible.”

He also wrote a FAQ for that campaign. And, since I already wrote an article controverting his numbers, I thought it would be more fun to do a “FAQ” of my own.

Enjoy!


Hey, I heard that Spotify is taking my subscription money and giving it to Top 40 artists. Are they counting my streams as streams for Katy Perry or something?

No, they absolutely are not.

You may get this impression if you didn’t get any further than the tl;dr version. But it’s not what is happening – and it’s not what Laguana is claiming either.

So, how is Spotify funneling my money to Top 40 artists?

That’s not exactly what’s happening either. If you listen to a track, Spotify pays royalties to the artists you listen to, and only the artists you listen to.

Well, what’s the deal then?

The issue is how Spotify calculates their per-stream royalties.

As I’m sure you’re aware, Spotify is a “flat-fee” service. Users pay a flat fee ($9.99 as of this writing), and use the service as much or as little as they like. Naturally, this means some users stream fewer tracks than average, and other users stream more tracks than average.

Spotify takes all of the subscription money, and puts it into a single “pool.” 70% of this pool goes to the various copyright holders, and 30% goes to Spotify. This means that 70% of your flat fee (we’ll round this to $7.00) goes to copyright holders, and…

Is he saying Spotify should take less than 30%?

No, he is not. Now stop interrupting!

Don’t be a dick, it was a fair question.

You’re right, I’m sorry.

So, what about per-stream royalties?

Currently, Spotify takes that “pool” of income, earmarked for copyright holders, and simply divides it by the total number of streams on the service. The result is considered the “per-stream” royalty.

Other people have called this the “pro rata” model, so I’ll use that term too.

Obviously, this royalty is not fixed. The total pool fluctuates according to the number of subscribers. But even with the same pool, the per-stream royalty fluctuates according to the total number of streams.

When you do the math, this means that Spotify users who stream fewer tracks than average drive the royalty rate up. Conversely, Spotify users who stream more tracks than average drive the royalty rate down.

So I should listen to less music, to drive the per-stream royalty rate up?

That is not what Laguana is saying. (It is a logical conclusion to his argument, but never mind.)

In fact, if you deliberately listen to less music, then according to Laguana’s logic, you would be “ripped off” more than you are right now.

Huh?

It gets back to the variable number of streams per user. In effect, the lighter users are “subsidizing” the per-stream royalties of the heavier users.

If you are a Spotify user who streams fewer tracks than average, then you will increase the overall royalty rate. But you will also stream fewer tracks (obviously). The end result is that the royalties that go to the artists you listen to, will be less than the $7.00 generated by your subscription fee.

So, according to Laguana’s logic, Spotify is “ripping you off.”

If you are a Spotify user who streams more tracks than average, then you will decrease the overall royalty rate, but you will also stream a greater number of tracks. The end result is that the royalties that go to the artists you listen to, will be greater than the $7.00 generated by your subscription fee.

According to Laguana’s logic, you are “ripping off” Spotify.

Whoa, harsh language.

I’m just using exactly the same language that Laguana is using. If you think it’s harsh – and I would agree – then you should politely ask Laguana to stop using it, too.

Did Laguana actually say heavier listeners are ripping off Spotify?

Not in so many words – but that’s only because he doesn’t even consider users who listen to more streams than average because they’re music lovers.

He has not ever considered that it is the heavier user who is more likely to proselytize – to put your music in their playlists, to tell their friends, etc. They are more likely to come out to see your show. They are more likely to buy merch, “golden tickets,” or other scarce goods. They are more likely to support you on Kickstarter or Sellaband. (Lighter users can also do these things, of course, but they’re less likely to as a group.)

What he has mentioned are people who engage in bad behavior. He talks about “click fraudsters,” who use automated tools to artificially inflate stream counts (in violation of Spotify’s End User Agreement).

He also talks about the gym that he goes to, which (according to him) uses Spotify for background music twelve hours a day. This, of course, is also in violation of Spotify’s End User Agreement. (Spotify has started a service called Soundtrack Your Band to offer business licenses, but it’s not yet available outside of Sweden.) He also fails to mention that these businesses pay additional royalties to PRO’s like ASCAP and BMI – but never mind.

He has described heavy listeners as “trust-funders [who] have nothing better to do than kick out the jams all day long,” and his only example of a heavy listener is “one guy clicking on an artist 10,001 times.” (Don’t worry, dear reader, because he believes that “you” are one of the “10,000 fans who stream a song once.”)

He has also said that artists who benefit from these heavier listeners are “benificiaries of a slanted system,” and compared them to people who engage in bribery.

O-kay…

Right.

So is Spotify the only business that runs this way?

Not even remotely. As far as I know, all streaming services use the pro rata model.

It’s also somewhat similar to how PRO’s like ASCAP and BMI pay out royalties. The performance royalties are pooled, and doled out according to which songs are the most popular by region. (Of course, they don’t have the atomic data that Spotify has. They rely on sampling the bigger radio stations, or auditing the larger venues – which means that they are much more biased towards popular acts. But that’s a story for another day.)

The “pro rata” model is also incredibly common in the wider world. Some examples:

  • Libraries. Readers pay a single fee (through taxes if it’s a public library, fees if it’s a university library, etc.) but the library orders materials according to what is checked out the most.
  • Subway passes. Riders pay a flat fee for a month-long pass, and use the subway as much or as little as they like.
  • Lunch buffets. Diners pay a flat fee for an “all-you-can-eat” buffet.

I’m sure you can come up with more.

In all of these cases, light users end up subsidizing the costs of heavier users. And the heavier users have much more of an impact on spending policies (books ordered, subway resource allocation, ingredients ordered) than light users. But since the light users are still getting their money’s worth, nobody considers it unfair.

So what does Laguana want?

He advocates for a different method of calculating per-stream royalties. This method is commonly called the “subscriber share” model.

Instead of putting all the subscriber revenue into a big pool, it is kept separate. The per-stream rate is found by dividing that user’s subscription revenue by the number of streams that user listened to.

Under this model, your $7.00 is used to calculate the per-stream royalty of only those artists you listened to, and nobody else. Then, just like the pro rata model, the artists that you stream (and only those artists) are paid your custom per-stream royalty.

That actually sounds pretty fair.

It does, until you realize what the effects would be.

For Spotify users who stream fewer tracks than average, the per-stream royalty rate will be higher than it is now. These artists will make more money per stream than under the pro rata model.

For Spotify users who stream greater tracks than average, the per-stream royalty rate will be lower than it is now. These artists will make less money per stream than under the pro rata model.

So, if you’re an artist whose fan base consists of heavy users, you will do worse under Laguana’s proposal. The only artists who will benefit are those whose fan base consists of light users.

So it will penalize artists whose fans are music lovers, and reward artists whose fans are casual listeners?

Exactly.

Which artists are which?

It’s very hard to generalize about these things, but the “polyvore/monovore model” seems to be the most accurate. (Laguana does not believe this model is accurate, but I don’t know why.)

In this model, if a user listens to more music, then it will be music from a greater variety of artists (they are polyvores). The wider the variety of music, the more likely this music will come from artists who are less popular. A polyvore will certainly listen to Top 40 music, but they’ll listen to other music as well.

Conversely, if a user listens to less music, then it will be music from a smaller variety of artists (they are monovores). Since they listen to a narrower variety of music, it is more likely to be music by popular artists. Monovores don’t listen to much music that’s not in the Top 40, and when they do, it’s only on rare occasions.

Note that I do not mean Top 40 pop music, though this may be the case. Both monovores and polyvores could listen to nothing but metal, and both will listen to Metallica. But only a polyvore would listen to Butchers of the Final Frontier or Abnormity.

Because the omnivores generate higher per-stream royalty rates than polyvores, the subscriber share model ends up generating more revenue for popular artists (like Metallica) and less money for less-popular artists (like Abnormity).

So the rich get richer, and the poor get poorer?

Yes.

This is fair?

That depends upon your point of view.

Are you, casual listener, offended that you are driving up the royalty rate for bands you don’t listen to? Are you, music lover, offended that you’re driving down the royalties for popular bands?

Do you not give a shit about artists that aren’t in the Top 40 of whatever genre you like?

If you answered “yes” to these questions, then the subscriber share model is fair. Otherwise the pro rata model is fair.

Personally, I like artists who aren’t popular artists. I want them to make more money from Spotify and services like it. And I don’t really care that casual listeners are “subsidizing” their royalties, just like I don’t care that other people ride the subway more than me.

So I do not think it’s fair. Not even remotely.

So I guess I shouldn’t do this Silent September thing?

On the contrary. You should do that.

It is true that participants will be driving down the per-stream royalty rate for everyone on Spotify, including the artists they are silently streaming.

This would be a problem if there were enough participants to affect the per-stream rate. There aren’t. Spotify has over 20 million subscribers. It would take a few million participants to drive down the per-stream royalty rate. Especially to the point where major labels would even notice the change. To major labels, it’s a rounding error.

But, thankfully, the indie musicians will get roughly the same per-stream royalties they get now. To less-popular artists, that “rounding error” will be a lot more money.

In September.

After that it’s business as usual.

Advertisements

45 thoughts on “A “FAQ” About Sharky Laguana and Silent September

  1. LOL!!! Well now I know why your mind was closed: you had already written your post. Sigh.

    Fact check:
    Libraries are NOT pro rata: authors and publishers are paid for the purchase of every book. Employees are paid by the hour. No one is paid per “read”. If a book is checked out a lot the library may buy more copies of the book (but most likely it will just be one or two copies otherwise whole sections of the library would be devoted to 2 or 3 authors), and the author will be paid again when the book is purchased, not based on how many times the book is read.
    You can’t compare physical books to streaming music: in economics physical books are a “private good”(excludable: you can’t have the book unless you pay for it, and rivalrous: only one person can possess a book at a time). Streaming music is a “club good” (excludable: got to pay the service, but non-rivalrous: songs can be played by many people simultaneously).
    https://en.m.wikipedia.org/wiki/Good_(economics)

    Subways are NOT pro rata. Employees are paid an hourly wage or salary no matter how many people are riding the train. They are not paid per-ride-given. The owner of the subway (typically a government) sets the price and receives 100% of the money. Most riders pay per use but when riders pay subscription-style (via fast passes) the owner of the mass-transit system gets 100% of the riders subscription fee.

    Lunch buffets are not pro-rata either. Food vendors are paid for the food upfront at a price they set. They are not paid “per-meal”. Employees are, by law, paid a salary or by the hour. They are not paid “per-meal” either.
    The buffet price is set by the owner of the food, and they get 100% of the money paid by diners.

    If Spotify was run like a buffet they would either pay for songs outright (like food vendors) or pay a salary (like employees).

    There are no clear analogs to streaming music. We’ve never had anything like this before. The closest analogy we have at the moment is Netflix. Movies are not typically consumed countless times by the same subscriber so it’s not a perfect fit, but it is a subscription streaming content business. So how do they do it? They *license* movies and tv shows for a defined period of time (paying a onetime license fee upfront) or they buy them outright by making them from scratch.

    You mistakenly believe that because your experience as a consumer is similar than that means the business is analogous. This is clearly not the case. This lack of careful consideration can be found elsewhere in your post (example: you say that light users should take comfort in raising the royalty rate for their chosen artist [an effect that is dramatically more pronounced in subscriber share btw] but then go on to say that it will take “millions” of people participating in Silent September to noticeably change the royalty rate).
    Here’s an idea: get your calculator out and actually figure out how many people it will take to have a 5% impact on major label revenues in the U.S. (royalties are calculated regionally due to various and varying compulsory mechanical rates – another detail that you haven’t considered). I think you will be surprised. Hint: it’s not millions.

    Unlike you I put months of work and research into what I wrote. This is probably one of the reasons why academics who have researched the issue using real data agree with my take and not yours, and why there is an emerging consensus among well respected industry experts that subscriber share is a fairer way to distribute royalties.
    I did not start with a conclusion (as you apparently have) and I still have an open mind on the topic. I don’t have all the data (the services are not sharing it with the public or researchers) and it’s quite possible that there are areas where I am wrong. But it’s the degree of manipulation out there (which is ultimately theft from small artists who lack the resources to manipulate in return) that has persuaded me to get up off my ass and try and do something to make the system work better for small artists. My approach is to try and turn the manipulation up so that it is visible to all. Will it work? I don’t know. No one has ever tried before. Hard to say how effective it will be. But it’s better than just sitting around whining and doing nothing.
    You disagree this will help small artists, but you clearly haven’t done much work on the subject so I’m not sure why anyone would assign much weight to your views. Nonetheless, keep digging, maybe you will find something that will help small artists, and that would be great in my view. I don’t have any special need to be right, and I look forward to this question being definitively answered so I can do something else with my time.

    Like

  2. My earlier comment didn’t post. That’s a bummer if it got lost when I hit send. Your examples are not pro rata/big pool. Not even close.

    As for it taking “a few million participants to drive down the per-stream royalty rate”. Nope, you are way off base there too.

    Labels will start noticing somewhere long before we get to a million participants. In fact I think they may start to notice as early as 15,000 participants. See the math that establishes this in question #7 in the FAQ.

    View story at Medium.com

    Like

    • As for it taking “a few million participants to drive down the per-stream royalty rate”. Nope, you are way off base there too.

      Well, technically, it’s not how many people participate. It’s the increase in the number of streams by those who do.

      What is the total number of streams required to lower the royalty rate by 5%?

      Well, according to your numbers, there were 4.7 million subscribers, and they generated $25.7 million in revenue. They also listened to 3.4 billion streams. (By the way, I’d like to know where you got those numbers; you don’t actually cite any sources in your FAQ.)

      This means that the per-stream royalty rate is $0.0075588235294118. Now, let’s say the goal is a 5% decrease in royalties. The royalty “goal” is thus $0.0.0071808823529412. How many additional streams would it take to reach this goal?

      The total number of streams needed is 3,578,947,368 streams (almost 3.6 billion). The difference is 178,947,368 streams – about 180 million.

      In order for 70,000 participants to reach this goal, they would, on average, have to listen to an additional 5,397 streams per month – 85.2 extra streams per day. The average song length in 2010 was 3.91 minutes. So, that’s an extra five and a half hours per day, each, beyond what they already listen to.

      That’s not gonna happen. Especially since, I suspect, most people who care about this stuff are already listening to more music than average.

      And that’s presuming there will be a constant number of subscribers. There won’t be. In 2014 alone, Spotify signed up 10 million subscribers – doubling their total subscribers from 2013. If Spotify gains one million more U.S. subscribers in 2015, those new subscribers will more than offset your 70,000 participants.

      So, if your goal is to intentionally be a problem, it’s not going to work.

      But, as I said, go for it. The worst that will happen is that indie artists make more money.

      Like

      • Aha, I didn’t see the link to the Audiam data. (Perhaps it would be better if it were in an inline link.)

        That’s useful; I’ll go through it.

        Like

      • “In order for 70,000 participants to reach this goal, they would, on average, have to listen to an additional 5,397 streams per month – 85.2 extra streams per day. […] So, that’s an extra five and a half hours per day, each, beyond what they already listen to”

        If they are what I’m told is a typical subscriber (200 to 300 streams per month) then they are listening to 10-20 hours a month depending on their favorite genre (see http://daleswanson.blogspot.com/2013/08/genre-average-song-lengths.html). That’s 20-40 minutes a day.

        So if they stream 24 hours a day, 7 days a week (by leaving the music on repeat) then they will stream anywhere from 10,800 to 14,400 streams. Subtract that 200 to 300 they are already doing and we get a boost from 10,500 to 14,200.

        There are all kinds of things that could stop this from happening. For example services could put in code to stop repeats after a set amount of time. But there’s also all kinds of things that can push it the other way: for example our house has Sonos, and we often play different music in different rooms *from the same service*.

        “Now, let’s say the goal is a 5% decrease in royalties”

        The goal would not be a 5% decrease in *all* royalties, just the major labels royalties, which makes it much easier. This is a multi-step process which looks horrible in text, so I’ll just paste a link to the spreadsheet. You can copy the spreadsheet and play with the numbers/check my math.
        https://docs.google.com/spreadsheets/d/130FaFkO_BdBoN9vVBJkbs0PLMoJmQZkq8HXzzm8XyFM/edit?usp=sharing

        Like

  3. I had a more elaborate rebuttal, but this will have to do:

    Libraries: Employees get paid by the hour, or by salary. Authors are paid per book sold. Neither are paid “per book read”, or “per book checked out”.
    Subway passes: Subways are owned by a single entity (typically the local goverment). They get 100% of the subscription revenue. Employees are paid by the hour or by salary. No one is paid “per ride taken”.
    Lunch buffets: buffets are owned by a single entity (the restaraunt). The restaraunt gets 100% of the subscription revenue. Food vendors are paid outright for the food. Employees are paid by the hour or by salary. The restaraunt keeps whatever is leftover after expenses. No one is paid “per meal eaten”.
    Netflix: Netflix licenses content for finite periods of time. The content is owned by a single entity (typically a TV or Film studio). The owners of the content get an up-front license fee which they negotiate with Netflix. No one is paid per-stream.
    http://www.quora.com/Netflix-How-does-streaming-licensing-cost-work-Is-it-per-play-per-user-What-is-the-approximate-range-of-per-play-cost-in-dollars

    Now let’s look at music:

    Music: Music is owned by a single entity (either the label, or the artist, or a combination of both). Spotify does not own or manufacture music. They are just the platform. In exchange for Spotify being the platform they get 30%.
    If streaming was like libraries then music would be purchased outright for use on the streaming platform. I would have no problem with this.
    If streaming was like Netflix then spotify would license music for a fixed fee for a finite time period. I would have no problem with this.
    Streaming cannot be like subway passes or lunch buffets because there is more than one piece of music on the network, so there is more than one owner. In fact there’s millions of owners. However, to the extent that streaming can be like subway passes or lunch buffets, Subscriber Share does a better job because it’s giving each owner it’s proportionate share of each subscriber’s plays.

    You don’t understand what pro-rata royalty distribution is yet. Keep working on it.

    Like

    • You’re totally missing the point here. My goal was not to make some kind of “perfect analogy” between Spotify and the other “pro rata” services.

      The point is that, in all “pro rata” systems, two things happen:
      1. Lighter users “subsidize” heavier users;
      2. Heavier users have a disproportionate effect on the operation of the service, since they use it more.

      But none of the users think this is unfair, because even the lighter users are getting their money’s worth. Even though both of these things are true, nobody is getting “ripped off.”

      Like

      • “The point is that, in all “pro rata” systems, two things happen:”

        Dude, no.

        In the context of Streaming Royalties Pro Rata is referring to the royalty distribution method. It’s latin for “in proportion”. Pro rata could also fairly be used to describe Subscriber Share, but it’s been adopted as the term of choice in academic/industry circles for the current royalty system. (side note: I opted for “Big Pool” as I think that’s more descriptive, and less jargon-y [side-side note: at least one academic has said he prefers my term “Subscriber Share” as opposed to the term he previously used “User-Centric”]).
        Pro-rata has *nothing* to do with lighter users subsidizing heavier users. You won’t find the word “subsidize” anywhere in the wikipedia listing for example:
        https://en.wikipedia.org/wiki/Pro_rata

        Just take the extra five minutes before you hit post to see if it’s actually true. Google is your friend.

        “But none of the users think this is unfair”

        Actually many users think this is unfair. Particularly users with esoteric tastes. Do you want a list? It’s a long, long, long list. This guy made a video (with animations!) he thought it was so unfair: https://www.youtube.com/watch?v=iCGNhT5fI5I

        “because even the lighter users are getting their money’s worth”

        Maybe in your universe lighter users don’t give a shit about anything other than whether there’s tunes coming through the streaming thingamajig. But I know many people who expect that when they pay for music that their money will actually support the musicians they listen to. Not for altruistic reasons (although some have those), but for a simple selfish reason: if you financially support music you love, you stand a better chance of getting more music from those people.

        Like

  4. Hey, Sharky.

    For some reason, your posts are still waiting for admin approval in the WordPress system. Still not sure why that’s happening (I double-checked the settings so you should be approved automatically). It could be because you provided different emails; but that’s just a guess.

    I may write WordPress about this to see what’s going on. In the meantime, I’m approving your comments when I can.

    Anyway, rather than reply to multiple comments, I thought I’d “collapse the thread” and reply to all your recent comments in one go.

    Now, then. About the campaign:

    If they are what I’m told is a typical subscriber (200 to 300 streams per month) then they are listening to 10-20 hours a month depending on their favorite genre (see http://daleswanson.blogspot.com/2013/08/genre-average-song-lengths.html). That’s 20-40 minutes a day.

    So if they stream 24 hours a day, 7 days a week (by leaving the music on repeat) then they will stream anywhere from 10,800 to 14,400 streams. Subtract that 200 to 300 they are already doing and we get a boost from 10,500 to 14,200.

    Well, the link you posted broke down song lengths by genre. I was going from this graph:
    https://plot.ly/~RhettAllain/131/average-song-length/

    If we use something other than average song length, then we have to make some pretty big assumptions about which genre a “typical subscriber” listens to. That takes us down a whole new rabbit hole, and I’m not sure it’s completely relevant. (If it is, then you also have to consider which genres a “typical” participant in the campaign listens to.)

    By the way, your “streams per month” number seems incredibly low to me, and I’m still not sure where you get that number. Spotify itself says each active user listens to 110 minutes of music a day, on average. The Audiam data you linked to shows that, for Spotify premium users in 2014, an average listener streamed about 722 tracks per month. I’ve yet to see any source for the 200-300 streams per month.

    But, anyway. Your numbers assume that the average participant will stream music all day every day for the entire month. There’s no earthly way significant numbers of people are going to do that, even if they do participate in “Silent September.” And there’s no way for the “all day every day” participants to offest those who don’t (there are only 720 total hours in September). It is statistically impossible to get that as an average.

    Also, you’re assuming that participants listen, on average, to 200-300 streams per month. I really doubt that. Both your readers and the participants are self-selected, so they almost certainly won’t be an average listener. The kind of listener with that low of a stream volume, probably doesn’t care enough about the music business to have even heard of either of us.

    I’ll just paste a link to the spreadsheet. You can copy the spreadsheet and play with the numbers/check my math.
    https://docs.google.com/spreadsheets/d/130FaFkO_BdBoN9vVBJkbs0PLMoJmQZkq8HXzzm8XyFM/edit?usp=sharing

    Your math is wrong. For example, you ended up “double-dipping” when calculating the majors’ per-stream rate. It will always be exactly the same as the indies’ per-stream rate (under the “big pool” method).

    You have to take the number of major-label streams multiplied by the “non-Silent-September” per-stream rate, and compare it to the same number of streams multiplied by the “Silent-September” per-stream rate. (That assumes that no participants accidentally stream a major label track, of course.)

    To get that number of streams, you have to multiply the total streams by the major-label market share. This is not quite as easy to do, and I’m pretty sure your numbers are off.

    First, the market share differs between physical vs. digital, and presumably between streaming vs. downloads. Second, the market share is different for publishing and for sound recording copyrights – so even the same song may not be totally “indie” or “major label.” However, the lion’s share of royalties go to sound recording copyright holders, so it’s probably best to use that percentage.

    I used Ovum’s data for this:
    https://musicandcopyright.wordpress.com/2015/04/28/recorded-music-market-share-gains-for-wmg-in-2014-sonyatv-is-the-publishing-leader/

    Their number is 76.1% for major label sound recording copyright holders, for digital revenues, from 2014.

    Of course, if the protest gathers enough steam, that market rate will not stay the same. It will decline, at least on the service itself.

    I’ve copied that spreadsheet and updated it to reflect these changes. I also put in a slightly more realistic figure: 20,000 participants, listening and additional eight hours per day on average. (To be frank, I’d be surprised if it even went this high.)

    With those numbers, the decline in royalties for major labels would be 2.10%. Luckily, the increase for indies is a more-noticeable 6.26%.

    To get a 5% decrease, you either need more protesters or more additional listening hours per day. For 17,000 protesters to decrease major label earnings by 5%, every single one would have to listen to an additional 23.5 hours of music per day, all day, all month.

    EDIT: Oops, here’s the spreadsheet:
    https://docs.google.com/spreadsheets/d/1OKwn4iRUCBRQpf-pnl_jGHTM1Pi-0vwAzx9YxCFT-pM/edit#gid=0

    “The point is that, in all “pro rata” systems, two things happen:”

    Dude, no.

    I know Spotify is not a “pro rata” billing model. But it’s what people are using to describe it, so I used it too. (A real “pro rata” model would be something like an electric bill, where users pay more or less according to how much they use the service.) That’s why I put it in “scare quotes.”

    If you want to call it the “big pool” model, or something, then that’s fine. But the examples I listed are also “big pool” (or whatever) models.

    Fact check:
    Libraries are NOT pro rata: authors and publishers are paid for the purchase of every book. […] Subways are NOT pro rata. Employees are paid an hourly wage or salary no matter how many people are riding the train. […] Lunch buffets are not pro-rata either. Food vendors are paid for the food upfront at a price they set. They are not paid “per-meal”.

    You would have a point if the “big pool” had no effect whatsoever on payouts. This is simply untrue.

    Libraries decide how many books to order according to how many times those books are checked out by the “big pool” of readers. Subway systems make scheduling, maintenance, and planning decisions – including hiring and wage decisions, both of employees and outside contractors – from usage by the “big pool” of riders. Restaurants with lunch buffets determine what to buy from food vendors according to the “big pool” of meals eaten.

    The mechanism by which those payouts are made (per book, per hour worked, per case of carrots, whatever) is beside the point.

    You mistakenly believe that because your experience as a consumer is similar than that means the business is analogous.

    Your campaign is explicitly targeting consumers. Since they are the ones you’re trying to convince, it is their perception that is all-important for this particular campaign.

    (Also, I’m not just a music “consumer,” but never mind – it’s not important.)

    This lack of careful consideration can be found elsewhere in your post (example: you say that light users should take comfort in raising the royalty rate for their chosen artist [an effect that is dramatically more pronounced in subscriber share btw] but then go on to say that it will take “millions” of people participating in Silent September to noticeably change the royalty rate).

    First, I did not claim that “light users should take comfort in raising the royalty rate for their chosen artist.” I said – accurately – that light users raise the royalty rate for all artists, whether a “chosen artist” or not.

    And, likewise, heavy users (including Silent September participants) lower the royalty rate for all artists, whether a “chosen artist” (including silently-streamed artist) or not.

    That is completely separate from quantifying how much they raise or lower the royalty rate. And, yes, for a streaming service with tens of millions of users, it will probably take one or two million users changing their listening behavior to have any effect that is more than negligible. At least in the long term.

    Unlike you I put months of work and research into what I wrote. This is probably one of the reasons why academics who have researched the issue using real data agree with my take and not yours

    Really? Because the two Scandanavian studies that you have linked to, both agree with me: the benefits are slight, but the artists who will benefit are the artists who are already popular. Everyone else fares (slightly) worse.

    I did not start with a conclusion (as you apparently have) and I still have an open mind on the topic.

    Uh, if you look at your previous posts and comments on this subject, you absolutely started with a conclusion (back in the comments on the Hypebot article). And you’ve done nothing but double down on it, despite many people (not just myself) questioning both your numbers, and your general premises.

    I disagree with you because I did crunch the numbers, using whatever data I could find, and came to a different conclusion. (One that absolutely seemed obvious to me, I will admit.)

    I do agree that the only way to definitively show anything is to have complete access to the raw data, and that’s not forthcoming. Audiam’s is the closest; but it doesn’t measure listening trends in users, i.e. the curve of light vs. heavy users, and which artists they tend to listen to. The two Scandanavian studies did crunch the numbers on those trends, and support the “polyvore/omnivore” model (as Touve called it) – the one that you don’t accept.

    Given the data that I have (including data that you have shared), your conclusions are simply not correct. And to create an entire protest campaign, driven by outrage over those incorrect conclusions, does not reflect someone with “an open mind on the topic.”

    IMHO, of course.

    Like

    • “an average listener streamed about 722 tracks per month. I’ve yet to see any source for the 200-300 streams per month”

      Average is not typical. Read my article. Or pick up a math book. The average is always above the median in positive skewed distribution sets.
      Source: employees at 3 different services said “200-300 streams per month” was most common usage. Outliers raise the average. None of the employees could speak on the record. It’s up to you whether you want to believe me or not but I have no reason to lie.

      “Your numbers assume that the average participant will stream music all day every day for the entire month. There’s no earthly way significant numbers of people are going to do that”

      I’m doing it. Can’t tell who else is.

      “The kind of listener with that low of a stream volume, probably doesn’t care enough about the music business to have even heard of either of us”

      That’s roughly my streaming volume. I certainly don’t listen to an average of 2 hours a day.

      “Your math is wrong. For example, you ended up “double-dipping” when calculating the majors’ per-stream rate. It will always be exactly the same as the indies’ per-stream rate (under the “big pool” method).”

      It is the same per-stream for both:: $0.0072645. You are simply wrong.

      “Their number is 76.1% for major label sound recording copyright holders, for digital revenues, from 2014.”

      I used 64%. Plug in 76% if you like. That will only make the effect more pronounced.

      Like

      • “A real “pro rata” model would be something like an electric bill, where users pay more or less according to how much they use the service.”

        No.

        “The mechanism by which those payouts are made (per book, per hour worked, per case of carrots, whatever) is beside the point.”

        Wrong, the mechanism by which people are paid is the only point!!!!

        Like

      • Wrong, the mechanism by which people are paid is the only point!!!!

        No, the point is that light consumers will always have less of an impact than heavy consumers. That’s because resources are allocated by aggregate use rather than per consumer.

        Don’t think that’s fair to light users? Then you should protest every other business that charges the same fee regardless of use. There are a ton of them, so you’ve got a lot of work to do.

        In all honesty, that would be the most “fair” system of all. Charge different users different prices according to how much they use the service. Charge light users less than they are being charged now, and heavy users more than they are being charged now. Tiered plans, usage caps, and no “unlimited” anything.

        Of course, you know what would happen: hardly anyone would sign up, and many users would drop out. Those sorts of plans are extraordinarily unpopular with consumers. But even if nobody dropped out, it would still result in less money. The heavy users wouldn’t pay more, they’d just stop listening when they hit the cap. I’m sure the light users would love to pay half of what they’re paying now, though.

        Short of that, there’s no real way to make it “fair.” Either light listeners subsidize the payouts of the heavy listeners, or you penalize artists whose fans are heavy listeners and reward artists whose fans are light listeners. The latter ends up being worse for indie or niche artists, so I’m against it.

        Like

      • Quick reply:

        It is the same per-stream for both:: $0.0072645. You are simply wrong.

        Huh. I remember seeing that it was slightly different when I first looked at the spreadsheet. I could easily have been wrong, since the method you are using to calculate the major label royalties simply doesn’t make sense. You’re assuming the market share is constant, then multiplying it by the total royalties payable minus the additional “indie royalties” generated by the protest. But that’s not how streaming services actually calculate royalties.

        I assume you looked at my spreadsheet, and can see the formulas there. That literally represents how we agree “big pool” royalties are calculated. You can tweak it if you want; read the comments.

        And I still don’t buy that many people are going to listen to an extra 10,500 streams. If an average song is 3.91 minutes, that’s almost 23 additional hours per day – every day – for a month. This is on top of what people already listen to – so they can’t listen on their cell phones on the way to work, or while at work, or at school, or the gym, or what have you. They can do nothing but let Spotify play on their computers, with an indie playlist on constant repeat.

        You might get a couple dozen people doing that. I doubt it would be more. You’ll be lucky to get 20,000 protesters to simply put on Spotify while they’re sleeping, and that won’t get you even halfway there.

        To put it in perspective: however you want to break it down per user, your protest will have to generate over 180 million extra streams this month to decrease major label revenue by 5%.

        Like

      • Really? Because the two Scandanavian studies that you have linked to, both agree with me: the benefits are slight, but the artists who will benefit are the artists who are already popular. Everyone else fares (slightly) worse.

        The authors used their own twitter accounts to express agreement with and support of my take:

        BTW both Arnt and Rex have also expressed interest in participating in the seminar I hope to give at SXSW in 2016 on this topic.

        You’ve hit on an important point here (most likely unknowingly), so I want to emphasize it: artists who are popular tend to do better. I’m sure you meant “popularity” in the sense of pop stars, but popularity is a relative term. Even among the most niche-iest artists, some are more popular than others.

        As Rex Pedersen (one of the authors you are referencing above) once said to me:

        “One thing I generally find that people tend to forget when they discuss this is that our conceptions of what is popular are often wrong. Among the top 100 artists for this specific month I can spot several (mostly local/Danish) artists that are probably only just making a living from their music at this point in their career. They appear side by side with the international superstars we expect to find in this category (although the local artists are probably only there for a short period). So a redistribution towards ‘the top’ is not necessarily a disadvantage for middle layer artists.

        You are assuming popular=bad, popular=pop, and popular=lots of people. There’s no requirement for any of these in order for an artist to be popular. You simply need to be looked upon favorably within a group. The group can be very tiny. For instance a student could be popular at a school that has only 10 students.

        We want to reward popularity – that’s a GOOD thing, and both models purport to do so. Big Pool tries to do this by looking exclusively at an artist’s share of global clicks. However this is prone to manipulation (indeed I am attempting to manipulate it myself!) and fails to give any consideration to the number of people who generated those clicks. This is a problem because the number of people who like something is the most important metric for measuring popularity. You wouldn’t say that someone who is liked a lot by one person is more popular than someone who is liked by a lot of people. Subscriber share takes a more nuanced approach, looking at market share of clicks among individual subscribers. Under subscriber share you need people to make money. You need to be popular with somebody.

        “Then you should protest every other business that charges the same fee regardless of use.”

        Subscriber share charges the same fee regardless of use, so it would be weird if I protested that. The only thing being changed is the royalty calculation method (i.e. how we distribute the proceeds), so any comparisons to other businesses should focus on how the proceeds are distributed.

        “Either light listeners subsidize the payouts of the heavy listeners,”

        Under Subscriber share light listeners will continue to subsidize heavy listeners in terms of providing the service. Both methods require all users to pay 30% to Spotify regardless of how much or how little they use the service.

        “or you penalize artists whose fans are heavy listeners and reward artists whose fans are light listeners”

        Penalty definition: “a disadvantage or handicap imposed on a player or team, typically for infringement of rules.”
        There is no penalty for artists with heavy listeners: they can earn up to 100% of what their listeners pay. If they want more money than 100% of what their listeners pay then they need to bring new subscribers into the system.
        It’s the same rule for artists with light listeners. There’s no disadvantage or handicap for either group: they can each earn up to 100% of what their listeners pay, and not a penny more.

        I cannot come up with any reason as to why an artist should be entitled to more than 100% of what their listeners pay.

        “The latter ends up being worse for indie or niche artists, so I’m against it.”

        Let’s examine what I think is your core assumption:

        Heavy users have more diverse listening habits.
        There was research in 2007 by Harvard professor Anita Elberse that supported this statement (but it was not necessarily compelling research – the Economist panned her work). This research only looked at individuals. It did not consider groups. Subscription streaming is in a *much* different place than it was in 2007, and is far more likely to be used in group settings (offices, gyms, hair salons, etc) than it was back then. The advent of wireless speaker systems like Sonos has really changed the game.
        I’m not aware of any research looking into music choices made by groups, but common sense says that groups will be less diverse than individuals. I’m reminded of this every time I get in the car with my family, and find that most of my music preferences are vetoed, or when I listen to my employees argue about what to play.
        Groups are also more likely than individuals to play music continuously, so they get more votes counted than individuals do.
        Finally, Elberse says that MOST users listen to something that only a few other people do. In other words, nearly everyone has an obscure artist they enjoy. Heavy users listen to more obscure artists on a quantitative level, but they also listen to far more popular artists too. So from a royalty perspective the heavy users not only cancel themselves out in terms of providing support for niche artists, they tilt the field towards artists with large numbers of fans, and away from emerging artists, who are the people I think are deeply disadvantaged by the current system.
        More recent research underscores this:
        https://people.jacobs.cornell.edu/mor/publications/thegoods/Park_ICWSM2015_MusicalDiversity.pdf
        “Our results show that volume and entropy might not be the best solution for computing the musical diversity of people on a highly complex map of musical categories such as subgenres.”

        Apologies, but I will now leave this conversation. Best of luck to you going forward.

        Like

      • The authors used their own twitter accounts to express agreement with and support of my take:

        The authors’ feelings towards you, personally, does not mean that their studies support your worldview, nor do they claim it does.

        Here is what those studies actually found:

        Among the top 5.000 artists, per user distribution would primarily benefit the most popular artists at the expense of the less popular. Top 1% among top 5.000 artists would go from 28,2% of payout with the current model to 31,0% of payout with the per user model. Artists between 1.000 and 5.000 would go from 18,1% of payout with the current model, to 15,9% of payout with the per user model – a relative decrease of 12,1%[.]

        – Music Streaming In Denmark

        The top 5000 artists this month would receive 89,7% of revenues in the user-centric model, compared to 89,2% in the current pro rata model – a small growth of 0,5 percentage points. In other words, the long tail of very small artists further down the list would receive a small amount less in a user-centric model, than in the current model, and both the top and medium artists would receive a bit more of the revenues in a user-centric model than the pro rata model.

        – User-centric settlement for music streaming

        Which is exactly what I’ve been saying.

        I’m sure you meant “popularity” in the sense of pop stars, but popularity is a relative term. Even among the most niche-iest artists, some are more popular than others. […] You are assuming popular=bad, popular=pop, and popular=lots of people.

        This is absolutely not what I meant, and I made it explicitly clear in the article: “Note that I do not mean Top 40 pop music, though this may be the case.” Don’t put words in my mouth.

        In fact, the one who has been playing the “Top 40 vs. Indie” card has been you. From the very beginning, you’ve presented it as “Avicii stealing from Butchers of the Final Frontier” or similar. This has not gone unnoticed among your fans: the video you linked to outright claimed “I just paid someone to listen to Nicleback.”

        We want to reward popularity – that’s a GOOD thing, and both models purport to do so.

        Here’s the difference. Under the current model, popular artists are proportionally rewarded – by how much users listen to them. Under the “subscriber share” model, popular artists are disproportionately rewarded. They are not rewarded by how much users listen to them, but by how much users don’t listen to anyone else.

        Penalty definition: “a disadvantage or handicap imposed on a player or team, typically for infringement of rules.”

        This is exactly what you’re doing. Here is your rule: “I cannot come up with any reason as to why an artist should be entitled to more than 100% of what their listeners pay.” You’re imposing a disadvantage upon less-popular artists for breaking this “rule.”

        Let’s examine what I think is your core assumption:

        Heavy users have more diverse listening habits.
        There was research in 2007 by Harvard professor Anita Elberse that supported this statement

        There is more research than that. From Music Streaming In Denmark, again:

        The number of streams per dedicated listener gives an indication of the intensity of use among listeners from different segments. The clear tendency among the top 5.000 artists is that the most popular artists have the least intensive listeners. Listeners of top 50 artists on average listen to 208,6 tracks per month, whereas less popular artists, on the other hand, are generally preferred by listeners that also listen to many tracks during a month.

        You bring up things like “group listening” and so forth, but you don’t actually provide any numbers for how much Spotify users engage in this “group listening.” You also make your own assumptions about it (e.g. that groups of people listen to less diverse music) – which may or may not be true, but are not backed up by any data.

        More recent research underscores this:
        https://people.jacobs.cornell.edu/mor/publications/thegoods/Park_ICWSM2015_MusicalDiversity.pdf

        That research paper is studying the relationship between listening diversity and socioeconomic status (race, gender, income level, etc). It did this solely by examining Twitter feeds from Last.fm users.

        It does not, and does not claim to, measure the relationship between listening diversity and hours spent listening to music.

        In fact, whatever your criticism of Elberse or other studies, you have provided absolutely no data that supports your assumption: that light listeners have more diverse listening habits.

        Apologies, but I will now leave this conversation. Best of luck to you going forward.

        Well, best of luck to you too.

        Like

      • The authors agree with me because I carefully read their work, and I’m not saying anything that contradicts it.

        You forgot to bold the first part of the sentence:

        Among the top 5.000 artists

        We’re already talking about the top artists here. Among the very top artists, artists that have already achieved success, there was a slight shift towards artists that were more popular.
        Again, this is a good thing, artists with more fans should be paid more than artists with less fans, as they are bringing in more revenue.
        But don’t be misled into thinking that this is a wealth transfer from the poor to the rich: you have to be very careful about drawing conclusions based on averages drawn from sample sets which contain a large number of outliers. For instance among those top 5,000 artists there were some big losers. They didn’t have that many fans. But they had a lot of clicks.
        Neither paper examined what happens at the long tail. The place where $100 is a factor in determining whether you have to sell your blood to buy a burrito (as I did on many occasion). But I can tell you what happens:
        As an average, the long tail gets less. There’s a lot of fraud in the long tail because it’s harder (I.e. Impossible) for the services to see at such a small scale. But don’t fret, this is a good thing because non-fraud artists who have real talent and have real fans will do dramatically better. Talented artists who previously auctioned off their uncashed spotify checks in Kickstarter campaigns won’t be inclined to do this anymore.
        The determinative factor is talent. If you can’t entertain real people you won’t do as well financially. It gets a lot harder to become cash flow positive on sheer manipulation of click numbers. You need paying subscribers, real people, to support you.
        Karl, it’s been fun, but you aren’t very careful with your analysis. I’m going to stop reading now. Wishing you the best.

        Like

      • You forgot to bold the first part of the sentence:

        Among the top 5.000 artists

        We’re already talking about the top artists here

        That’s because they didn’t even look at anyone not in the top 5,000. If they had, they would likely have found that the under-5000 would also have gotten less. Just like the other study (which, despite your assertion, did look at all artists) found.

        But don’t be misled into thinking that this is a wealth transfer from the poor to the rich

        Since popular artists are already doing better than unpopular artists, yes, it is. We can haggle about the definitions of “poor” and “rich,” but it is transferring money from artists who already make less right now, to artists who already make more right now.

        As an average, the long tail gets less. There’s a lot of fraud in the long tail because it’s harder (I.e. Impossible) for the services to see at such a small scale. But don’t fret, this is a good thing because non-fraud artists who have real talent and have real fans will do dramatically better.

        Any artist will only do better if their fans listen to much less music than average. Under your proposal, the determinative factor is not talent. It is not how well you entertain people. It is the listening behavior of your fanbase. You and I may (theoretically) have the same number of fans, and entertain them equally well. But if your fanbase listens to more music overall, then you’ll do worse than me.

        And are you seriously claiming that having fans who listen to more music makes you guilty of “fraud?” That these artists don’t have “real talent” and “real fans?”

        Personally, I think that’s absolute nonsense. People should listen to more music, and a wider variety of music. That’s good for musicians, and it’s good for society.

        Like

      • “But if your fanbase listens to more music overall, then you’ll do worse than me.”

        Nope. Not true at all. Let’s say we have a thousand fans each. And we are each wildly popular with our fans averaging 50% of the streams (this number doesn’t matter – what I’m about to say is just as true whether we each average 5% of the streams or 100% of the streams)

        Your fans are light users who only play 100 tracks a month.
        Under Big Pool you’ll get $350.
        Under Subscriber share you’ll get $3,500.

        My fans listen to 800 tracks a month on average. That’s a lot. That’s anywhere from 3 to 4 times what most people do.
        Under Big Pool I’ll get $2800.
        Under Subscriber share I will get $3,500.

        My fanbase listens to more music overall but I get the exact same payout you do because we each entertain our fans 5% of the time, making us eligible for 5% of the royalties.

        Note also: we both get more money, considerably more (you are getting 10x!), than we would get under the Big-Pool.

        See here if you want to see what this look like:
        https://docs.google.com/spreadsheets/d/1etVOcTWYIi61uWCN2zK_8wgP2ChuKxOZokGPtsLwjxA/edit?usp=sharing

        So who are the losers? Someone who can get a fanbase that manages to listen to music more than 1,000 streams per month. That’s not going to be easy to do, considering the following truths:

        1. Streaming usage follows a power law: as the number of streams increases, the number of users averaging that amount decreases. See: https://en.wikipedia.org/wiki/Power_law
        2. As a result of the power law, the graph for streaming usage is what statistics calls a positive skewed distribution set with a continuous variable.
        3. In a positive skewed set with a continuous variable, the average is always above the median (there’s an exception for multiple tails – but we don’t have multiple tails here, so we don’t have to worry about that).
        4. Result: “Above average” users are going to be a minority of all users.
        5. As the streams increase they become an even smaller group of people.

        Bottom line: Attracting a couple above average subscribers is not sufficient. You need your entire fan base to be above average. And then you need them to stream you a lot.

        If you can pull this off, you will not only make more money per fan than someone getting the same % of streams of below average fans, you’ll actually make more money than the fan paid!

        Turning a $10 fan into $500 of royalties? Awesome.

        This is what you are defending.

        OK I’m really done. Either you get it or you don’t. I’ve done what I can.

        Like

      • Nope. Not true at all. Let’s say we have a thousand fans each. And we are each wildly popular with our fans averaging 50% of the streams

        That is not what I meant at all, though perhaps I wasn’t clear.

        Let’s say we have a thousand fans each, and those fans listen to an equal number of streams of our music. Let’s call it 100 streams each per month. At this point, by any standard that is measurable on the streaming service, we are equally popular. Both our fans contributed an equal amount to the streaming service’s bottom line, we have equal numbers of fans, and both of our fans value our music equally. If you believe that talent or the ability to entertain people translates to fans (a big “if,” but never mind), then we are both equally talented and entertaining.

        Here’s the only difference. Your fans listen to 800 tracks a month on average, while my fans listen to 200 tracks a month on average.

        Under the “pro rata” model, we both earn an equal amount of money. But under the “subscriber share” model, I will earn four times as much as you do. Despite the fact that we have the same number of fans, and they listen to us an equal number of times. And it is determined, not by talent or entertainment value, but by the listening behaviour of our fans.

        See here if you want to see what this look like:

        Once again, your math is wrong. Very wrong. So wrong, I wonder how you could even consider that it is right.

        First of all, you do not calculate the per-stream royalty anywhere. You set the “pro-rata rate” and use this to calculate both subscriber share and “pro-rata” share. This is ridiculous, since it is the only thing that changes between the two rate calculations.

        Second, you assume that the artist’s share of streams is fixed between artists. As I explained above, this is not what I meant – and it’s absolutely not what happens in the real world, either. Users who stream more music don’t stream the same artists, in the same proportions, as users who stream less music.

        And I have no idea how you’ve tried to calculate the “What does it mean for the listener?” values. I guess you’re assuming that – for example – the “% of time spent listening to music, weekends only” value assumes that these listeners only listen to music on weekends, and never at any other time whatsoever? If that’s what you’re doing, it’s completely ridiculous. Nobody who pays for Spotify listens to absolutely no music during the week, and I double-dog dare you to find data that even hints at supporting this assumption. This entire section is complete nonsense.

        To make things more realistic, I’ve made my own spreadsheet, which corrects your errors:
        https://docs.google.com/spreadsheets/d/1qrPdMwoLWwmFdWdXRxVVKL1NDlp2lLglQuDbOk-eDbU/edit#gid=0

        Even though Artists A through F are equally popular, Artist A earns 180% more than she would under the current model, while Artist F earns 60% less. More importantly, Artist D – whose listeners stream exactly the average amount of streams – earns 30% less than she would right now.

        Look, I know that for some reason you’re wedded to this particular payout model. But it is not fair, it will not help atists who need it, and it rewards listening to less music. It is not good for either art or artistry, and I sincerely hope you admit you’re wrong on this one.

        Like

      • I am clearly incapable of not responding. :-(.

        It boils down to this: either you make every stream equal, or you make every fan equal. You can’t have both at the same time. They are mutually exclusive.

        You prefer every stream is equal. Your belief is that streaming volume is the best measure of popularity. However when you go that route this becomes possible:
        A con-artist who hires ONE worker to stream his catalog of 34 second songs 70,000 times in a month is going to be twice as popular as an honest band with 500 fans who each listen to a normal length album 7 times.
        As we are both aware, when this happens, not only is the con-artist getting paid for fraud, but every artist on the system makes less. There are incentives for pay-per-play beyond just short-term money too, such as artificially moving an artist up the charts.
        If you think this isn’t happening, you are living in a fantasy world:
        https://www.freelancer.com/job-search/spotify-plays/1/#8422745

        I believe every fan should be equal. I don’t think someone is worth more because they listen more, or worth less because they listen less. I think everyone is worth the same, because they all pay the same: $10.
        If we make every fan equal, then every user’s royalty distribution is capped at $7. There’s no financial motive to pursue fraud anymore, indeed it’s a money loser. So pay for play for money is broken, and artificial chart movement becomes a very expensive proposition.

        Hopefully you will agree with me that fraud, and artificial chart movement are bad for music?

        Your concern about popularity then moves towards your example: two people, A & B, who listen to a band 100 times. The band makes less from B if that user streams more than A.

        Let’s look at this from the listener perspective for a moment:
        If a listener pays $10/month, and listens to 200 tracks, then they are paying 5 cents a song.
        If a listener pays $10/month, and listens to 1,000 tracks, then they are paying 1 cent a song.

        Clearly the listener who listens less, is paying more on a per stream basis. This doesn’t bother you, nor should it. But for some reason it does bother you that an artist would get more per stream from a listener who is paying more per stream?

        I don’t share this concern. To me it makes perfect sense. And I think there’s a lot of good reasons we should go this route. First of all emerging bands will make a living wage much faster. Second of all it encourages artists to pursue light listeners and non-subscribers.

        But let’s dive a little deeper into the philosophy end of this: the big pool says there is a pie of money, and everyone gets a share of the pie based on clicks.
        Subscriber share says there is a pie of money, and everyone gets a share of the pie based on clicks.

        The only difference is whether that pie of money is global, or individual. So if there was only one user, there is no difference between Big Pool, and Subscriber Share. Likewise if everyone clicks the same amount there is no difference.

        There’s nothing morally wrong with either approach.

        So it boils down to which one is better for fans, artists, and the industry? I believe, and many people much smarter than me believe that subscriber share helps address some of the problems streaming currently suffers from (like fraud, chart manipulation, and the difficulty emerging artists face trying to get to a subsistence level).
        It also makes every user equally valuable, which makes sense to me, since every user is paying the same amount of money.

        To respond in brief to your points:

        “First of all, you do not calculate the per-stream royalty anywhere. You set the “pro-rata rate” and use this to calculate both subscriber share and “pro-rata” share. “

        You are mistaken. The Pro-rata rate of $0.007 has absolutely no role in calculating the subscriber share payout. For subscriber share we don’t need a per stream rate to calculate royalties, we only need to know what % of the time a subscriber spends listening to an artist. Then that artist gets that % of the $7 available.

        “Nobody who pays for Spotify listens to absolutely no music during the week”

        You are making a big assumption based on your own listening habits. Actually many people are very busy, and have jobs and kids and responsibilities that may prevent them from listening to music throughout the week.
        I should know, I am one of those people.

        To make things more realistic, I’ve made my own spreadsheet, which corrects your errors

        If you want other people to look at your spreadsheets, you have to make them public.

        Like

  5. By the way, it turns out that raw data is available. It’s simply a huge amount of raw data.

    That data is collected by Echo Nest, who provide Taste Profile data. Unfortunately for folks like you and I, it collects data from over 1 million listeners (not merely on Spotify), and over 300,000 tracks, and only provides raw data that links tracks and listeners. This means that the database has over 40 million entries.

    This is good for data wonks, but bad news for anyone who has a traditional spreadsheet program – which hangs after a million or so entries.

    I’m looking into software (R, custom scripts, whatever) that will be able to parse and evaluate this data, but don’t hold your breath.

    Like

  6. Karl I scoped this out a long time ago: “There is no connection to individuals. The date added field is the date of the anonymized ingestion and is not the date of the original activity.”
    All you get is a user ID, song ID, and a play count. So you can’t look at a specific user’s data for a specific month which is absolutely critical for calculating subscriber share vs big pool.

    It’s useless.

    Like

    • “There is no connection to individuals. The date added field is the date of the anonymized ingestion and is not the date of the original activity.”
      All you get is a user ID, song ID, and a play count. So you can’t look at a specific user’s data for a specific month which is absolutely critical for calculating subscriber share vs big pool.

      The ID’s are anonymized, meaning there is no connection to personally identifiable information (“usernames, listener details, original IDs, dates, IPs, locations”). But a single user will still have the same ID (as will a single song).

      The point about a lack of time frame is a good one, though. Without a time frame, it is impossible to tell if a user is a heavy listener, or if an Echo Nest partner has simply tracked their listening behavior for a longer period of time.

      That’s too bad. Back to the drawing board, I guess.

      Like

  7. Going back to your prior post, I see you are wanting a global-theory-of-everything spreadsheet that contains an entire universe of users and artists. That’s a big job! I’m not sure I can do that, but I’ll try and push the ball a little further down the road.

    One thing I notice about your spreadsheet is that it has what statisticians call a “normal distribution”. In other words it’s a perfect bell curve. You don’t find a lot of normal distributions in real world datasets, and there is zero chance that streaming usage will follow a normal distribution. Instead it will have a positive skew.

    See here for a description of normal distributions and positive/negative skews: http://onlinestatbook.com/2/introduction/distributions.html

    One of the hallmarks of a normal distribution is that the mean (aka average) and the median will be identical, which is true in your spreadsheet: 400. Another hallmark is that if you plot it out logarithmically it will form a perfectly straight line. Also true in your dataset.

    The reason streaming has a positive skew is that the high end of the range is not 700, as you have indicated. It’s at least 14,000 and is probably higher than that. But most users will be streaming far less. That long tail out to 14k is what makes it a positive skew. And this causes a “champagne glass” or “hockey stick” if you graph it out visually.

    To be a little more technical about it, I suspect streaming usage will follow a power law (see: https://en.wikipedia.org/wiki/Power_law) known as the Pareto principle (aka the 80-20 rule): https://en.wikipedia.org/wiki/Pareto_principle
    Roughly speaking, the top 20% of the users will be responsible for 80% of the streams.
    I would be very surprised if this is not the case.

    If you are approaching this with an open and fair mind then you definitely do not want to form an opinion based on looking at a normal distribution, and you want to consider a wide range of scenarios.

    So to try and present a fuller picture of these scenarios, I’ve created a new spreadsheet:
    https://docs.google.com/spreadsheets/d/1sbK_PS7t9ea9zl1WMi46W5AB_4nwS4-xM9rWRLEtPvQ/edit?usp=sharing

    This time I’ve followed your lead and created a little streaming universe. My universe has 100,000 subscribers who each pay $10 a month generating $1M in revenue. The average stream per month is 750, which is roughly what Spotify reported via section 115 disclosures (the actual average is 722 – but they also say elsewhere that the average user is streaming about 1,100 streams per month, see here, slide 11: http://www.slideshare.net/DigitalJourneys/digital-journeys-2014-understanding-music-consumption-in-the-digital-age, so I think rounding up slightly is fair).
    The Pro Rata payment per stream is .0093 which is almost identical to the actual Pro Rata rate for premium subscribers before Spotify started offering free 30 day trials (see Jan and February of 2014: https://docs.google.com/spreadsheets/d/1tQFoh6vOcWwMZoudquW-cEJI5eWE-EBI3jJLRJa1ixc/edit#gid=0)

    I divide these 100,000 subscribers into deciles based on usage. And I then look at 10 different artists, and compare the two royalty methods under a variety of scenarios.

    Sheet 1 follows the Pareto principle: 80% of the streams are generated by 20% of subscribers.
    Sheet 2 is a drill down of the top 10%
    Sheet 3 is a drill down of the top 1%
    Sheet 4 is a shallower skew, just for the sake of comparison.
    Sheet 5 is a normal distribution, just like you had on your spreadsheet.

    Each sheet has 10 artists. Each artist gets the same amount of overall plays. I have started this at 50,000 plays (but you can change this if you like – changing it on ). 50,000 plays at this scale is roughly equal to a top 20 artist on Spotify. We can use smaller numbers of plays but

    I’m going to ask for one consideration from you: it’s impossible to show individual subscribers and the spread of their individual streams without having an exponentially larger spreadsheet. So when we talk about the subscriber share let’s say, for the sake of clarity, that the % is equivalent to the # of fans. So if there are 10,000 subscribers, and the subscriber share % is 20%, that’s equivalent to 2,000 fans listening 100% of the time, and 8,000 fans not listening at all. We could break it down further but the effect is the same for the sake of royalties: a listener listening 100% of the time is the same as 2 listeners listening 50% of the time and 4 listeners listening 25% of the time, etc.

    So with all that being said, here’s what I see:

    1. Pro Rata puts elite listeners in control
    If the 80/20 rule is indeed how usage is distributed, then the top 10% controls a whopping 66% of the revenue! And the top .01% pays out $35k in royalties despite only contributing $700 to the pool.
    Even if we use the normal distribution the top 10% controls 4x as much royalties as the bottom 10%, despite paying the same amount of money.

    2. Pro Rata deeply penalizes artists for having fans who are not in the top 10%
    If you have the misfortune to have a fanbase made of light usage subscribers you are screwed. Artist D has attracted 667 fans, but all light users, so they make $466 under Pro Rata.
    Meanwhile Artist E has attracted just 10 fans, but they make $466 too. If Artist E had as many fans as Artist D they would earn over $31,000.
    But even if you do have heavy user fans you are still screwed. Artist H has increasing numbers of heavy usage fans but failed to get any of the top 10%. They are paid substantially less under Pro Rata.

    3. Subscriber share rewards you for generating revenue
    One of your concerns is that Subscriber Share “penalizes” artists for attracting heavy listeners. But it’s clear from this graph that Artists G and H would make more under Subscriber Share even though they have substantial heavy listeners. Artist C would make slightly less, but the difference is barely noticeable. The only big loser is Artist E, who’s fans are exclusively in the top 10%.
    I personally would find it suspect that an artist was able to attract only the top 10% and no one else, particularly when that’s the highest paying tier. But even assuming it’s legitimate, why do we want to give these artists 66% of all revenue, when they are only attracting 10% of the fans? What’s your reasoning here?
    This seems particularly weird when we consider that only 1% of the world has subscribed, and most of the unsubscribed are light users. Don’t we want to attract light users? Why are we penalizing artists for bringing the people we need the most?

    4. WHO are the top 10% anyways?
    Spotify says 23% of streaming happens at work, and 22% happens in the gym (see slide 10 in the link I provided earlier).
    At my gym the Sonos system is going non-stop 14-16 hours a day, playing mostly dance and hip hop. Yes it’s a violation of the ToS but it’s happening anyways.
    So is the top 10% made up of similar use cases? Again, why should this use case be valued so much more highly than a low usage subscriber who pays the same amount of money.

    Like

    • One thing I notice about your spreadsheet is that it has what statisticians call a “normal distribution”.

      Yeah, I remember my statistics course too.

      The problem with all of these models is that there is not a study that shows what the distribution actually is. It is almost certainly some sort of “bell curve,” but whether it is positively or negatively skewed (and by how much) is not something that is available with the current data.

      That’s why I was hopeful about the Echo Nest data – it seemed closest to providing the data for that kind of spreadsheet. Ah, well.

      To be a little more technical about it, I suspect streaming usage will follow a power law

      It is almost certainly not a power-law distribution. This would mean that almost all of the paying users listen to almost no music. The Pareto principle may work for certain things in business, but it does not work for services like this (or for lunch buffets, or transportation usage, etc).

      Instead, I’m betting that it follows a Poisson distribution of some kind. But both of these are just hunches without actual data.

      One interesting thing, however, is to look at the effect of Spotify’s user growth on the average streams per listener. Spotify doubled their number of paying subscribers in 2014. If you were correct, the majority of these users would be extraordinarily light users, and the average streams per user would decline significantly. But that’s not what happened. The average number of streams per user essentially stayed constant (at roughly 700 streams per month).

      So to try and present a fuller picture of these scenarios, I’ve created a new spreadsheet:

      I don’t have time now, but I’ll look it over after work. If it follows a power law, though, it’s not accurate. You’re choosing a distribution that proves your argument – it’s circular reasoning.

      One other thing:

      Spotify says 23% of streaming happens at work, and 22% happens in the gym (see slide 10 in the link I provided earlier).
      At my gym the Sonos system is going non-stop 14-16 hours a day, playing mostly dance and hip hop.

      This does not mean that Spotify is being played to the public at work or at the gym. These people are almost certainly listening on headphones while they sit at their desk or use the treadmill. My office, for example, does not allow music to be played out loud at all, but nearly everyone has headphones on while they work.

      More later.

      Like

      • “It is almost certainly some sort of “bell curve,””

        In statistics the term “bell curve” explicitly refers to normal/gaussian distribution.

        but whether it is positively or negatively skewed (and by how much) is not something that is available with the current data.

        Negatively skewed? You can’t have an average between 765 and 1100 and have a negative skew when a user can stream thousands, even tens of thousands, of tracks in a month. The only way it could have a negative skew is if small amounts of users streamed until we got to 700, then it swung upwards sharply, followed by a steep decline with no one streaming at all after 850 or so. Off the top of my head I can think of 4 examples within my personal zone of observation where they are streaming 10 hours a day, so this is clearly out the window.

        It’s a positive skew, no doubt about it.

        It is almost certainly not a power-law distribution.

        A second ago you didn’t know if it was even a positive or negative skew, now you are quite certain that it isn’t a power-law distribution. When you are uncertain as to the direction of the skew I am mystified as to how you could be certain what the skew looks like.

        This would mean that almost all of the paying users listen to almost no music.

        I don’t follow you.

        Here’s how one scientists describes power laws:
        “If the heights of an imaginary planet’s inhabitants followed a power law distribution, most creatures would be really short . . . [but] nobody would be surprised to see occasionally a hundred-foot-tall monster walking down the street.”

        As I imagine this applying to streaming: most users are comparatively short, between 200-350 streams per month. And no one is surprised to occasionally a hundred-foot-tall monster: someone who streams 5k, 10k, 15k streams a month or more.

        But since I am not a mathematician let’s go with a broader definition with less controversy: a fat-tailed distribution. Most users will stream some range under 750 streams per month, and as the number of streams goes up from there, the number of users who are able to stream that much gets smaller. This results in an overall average of 750. Can we agree on that? In other words, there should be no debate we are looking at a positive skew, the only question is the shape of the skew, and the width of the tail.

        “If you were correct, the majority of these users would be extraordinarily light users, and the average streams per user would decline significantly. But that’s not what happened. The average number of streams per user essentially stayed constant (at roughly 700 streams per month).”

        What if the ratio of heavy/light is the same in the general population? Adding more users would leave the average unchanged.
        We don’t know anything about behavior. What if users are using the service more as the service becomes better, more reliable, better bandwidth, etc.? Then the average could remain unchanged even if newer users were lighter (and I don’t follow you as to why newer users should be automatically lighter?).
        Finally I disagree that average streams per user has stayed “essentially constant”. I see a great deal of noise and variability. Swings of 115 streams per month are dramatic when we are talking about billions of streams. And when we layer in the fact that Spotify introduced free trails of premium services in 2014, I think it gets very difficult to say anything meaningful about trends.

        “You’re choosing a distribution that proves your argument – it’s circular reasoning.”

        Actually I chose three different usage distributions and modeled them all: power law, shallower skew, and normal. That should run the gamut. But if you want to see a different curve it’s easy to do. And for each distribution I modeled every kind of population distribution I could think of. If you can think of any I missed please let me know.

        “This does not mean that Spotify is being played to the public at work or at the gym. These people are almost certainly listening on headphones while they sit at their desk or use the treadmill. My office, for example, does not allow music to be played out loud at all, but nearly everyone has headphones on while they work.”

        Of course people listen privately in those occasions. But they listen publicly as well, and the public plays are more likely to be continuous since more than once person is being entertained. Are you suggesting that gyms and offices are never playing music openly? My experience is different.

        Like

      • Again, I only have a little bit of time to resopond. Expect a better response later.

        In statistics the term “bell curve” explicitly refers to normal/gaussian distribution.

        No, not the way most people understand it, at least.

        The term “bell curve” refers to some kind of distribution where there is a mode (slope = 0) and an “average” (normal) distribution (where the continuous distribution curve is divided such that 50% of the curve is below the average, and 50% is above it). Neither of these numbers represent the median (and no stats that I’ve ever encountered even mention the median).

        But in a “bell curve,” if the mode is greater than the average (a “right-skewed” distribution), the number of people who are below the average are greater than the mode (the “typical” user). On the other hand, if the mode is less than the average (a “left-skewed” distribution), the number of people who are below the average can be greater than the mode.

        There has been no study that I am aware of, that could possibly determine which model is more accurate. But, again, consider that of the users who signed up for Spotify, the average number of streams remained relatively constant – so whatever these users listened to, they listened with the same average intensity as the existing users.

        And no model is comparable to a “power law” model – which is not any kind of “bell curve” at all. It is simply dividing by two (or whatever constant) of the number of streams per listener as the number of listeners increases.

        Let’s say that you are right, and that the “average” user listens to 200 tracks per month. Under the power law model, the number of users that listen to 100 tracks per month is two times the number of listeners who listen to 200 tracks per month. And the number of users that listen to a mere 50 tracks per month is four times the number of listeners who listen to 200 tracks per month.

        As you can see, by the power law model, you will have over two hundred times as many listeners who listen to a only single track, as you will have listeners who listen to 200 tracks.

        This is completely ridiculous. First of all, there is a natural minimum number of streams, since anyone who goes below that point simply won’t sign up for the service. Second of all, there are only so many hours during the day that even the heaviest listener can consume, so there is an absolute maximum (which there isn’t under a “power law” distribution).

        Actually I chose three different usage distributions and modeled them all: power law, shallower skew, and normal.

        I will examine these. I have done my own work, and I will absolutely say with certainty that any kind of “bell curve” model (Poisson, normal, whatever) will result in more royalties for only those artists that have listeners who are lighter streamers than the mode. (It does not matter whether the mode is greater or less than the mean.)

        I can tell you one thing, though: given the uncontrovertible fact that lighter listeners listen to popular music, and that heavier listeners listen to less-popular music, then you will be transferring money from the artists who need it the most, to the artist who need it the least.

        All of the studies have determined that this is true, and no study ever has determined that this is false.

        Of course people listen privately in those occasions. But they listen publicly as well, and the public plays are more likely to be continuous since more than once person is being entertained.

        You have not shown one single shred of evidence that it is more than “once” person who is listening. (You claim that your gym plays Spotify, but even if that were true, it is one listener among dozens of millions.)

        Furthermore, if these actors were significant in any way, it likely would have shown up due to the fact that Spotify gained as many subscirbers this year as they ever had total. It does not.

        It also does not show up in the earlier studies out of Scandinavia – all of which showed that more-popular artists have fans that are lighter users than fans of less-popular artists. If indeed the numbers were skewed by listeners (like your gym) who were heavier listeners, but listened to more-popular artists, this would have been reflected in these numbers. It was not.

        There is absolutely no evidence that you (or anyone else) has presented that even suggests that these entities have any significant influence on streaming royalties. You are arguing against straw men, and straw men that have no material impact on the royalties we’re talking about.

        Are you suggesting that gyms and offices are never playing music openly?

        I’m sure many do. They simply have no material impact on the royalties paid out by streaming services. They are minority outliers that are more than made up for by individual listeners, who outnumber them by many millions to one. The fact that these individual listeners also happen to wear their private earbuds to the gym or the office does not make your argument any less insane.

        Like

      • In statistics the term “bell curve” explicitly refers to normal/gaussian distribution.
        No, not the way most people understand it, at least.

        Sigh. Look if you want to have a discussion, and I’m happy to do so, we need to use words the way are defined, not the way you think “most people understand them”.

        Let me help you…

        Bell curve definition:
        https://www.google.com/webhp?sourceid=chrome-instant&ion=1&espv=2&ie=UTF-8#q=bell%20curve%20definition

        Normal distribution definition:
        https://www.google.com/webhp?sourceid=chrome-instant&ion=1&espv=2&ie=UTF-8#safe=off&q=normal+distribution+definition

        OK now show me a picture of normal vs. skewed distribution:

        So let’s use these words the way are meant to be used OK? I can’t have an understandable conversation with you if you are making up definitions as you go along.

        There is no chance, none, zip, zero, that streaming usage will be a normal distribution (i.e. bell curve). The simple reason for this, is the graph will not be symmetrical. If you put the number of users on the Y axis, and the number of streams on the X axis it will be lopsided with a long tail extending to the right. It might have multiple peaks (i.e. multi-modal distribution), and it may offer some other surprises, but it will not be symmetrical.

        “But in a “bell curve,” if the mode is greater than the average (a “right-skewed” distribution), the number of people who are below the average are greater than the mode (the “typical” user). On the other hand, if the mode is less than the average (a “left-skewed” distribution), the number of people who are below the average can be greater than the mode.”

        This makes no sense at all in so many different ways.
        A positive (i.e. “right”) skewed distribution is NOT a bell curve. Bell curve means NO SKEW.
        See above.
        In a positive skewed distribution the mode is less than the average, not more, so you got your definitions backwards. See here for a refresher course:

        As you can plainly see, most users in a positive skewed distribution will be below average.

        “It is simply dividing by two (or whatever constant) of the number of streams per listener as the number of listeners increases.”

        You’ve got your constants reversed. What you are describing is simply an exponential function.
        Here is a graph of an exponential function:

        Here is a graph of a power law:
        https://en.wikipedia.org/wiki/Power_law#/media/File:Long_tail.svg

        Exponentials go to zero much faster than power laws. See here for a discussion the difference:
        http://math.stackexchange.com/questions/164436/difference-between-power-law-distribution-and-exponential-decay

        Like

      • Sigh. Look if you want to have a discussion, and I’m happy to do so, we need to use words the way are defined, not the way you think “most people understand them”.

        Yeah, that explanation was not the best – I was tired when I wrote it, and just wanted to go to sleep. Sorry.

        What I meant is that, for most people, a right-skewed or left-skewed distribution is also called a “bell curve,” even though neither is a normal distribution, because they’re vaguely shaped like “bells.” On the other hand, an exponential or power law distribution would not be a “bell curve,” since they do not have curves shaped like a “bell.”

        If we want to be specific, let’s just drop the term “bell curve” altogether.

        You’ve got your constants reversed. What you are describing is simply an exponential function.

        Cumulatively dividing by two will never result in a negative number (which exponential functions have). You probably mean an exponential distribution.

        And you’re right, I was thinking that you were talking about an exponential distribution. My bad.

        EDIT: But the numbers from an exponential distribution are actually much better than the numbers under a power distribution. If anything, I was being kind. See below for details.

        If you put the number of users on the Y axis, and the number of streams on the X axis it will be lopsided with a long tail extending to the right.

        This is actually backwards from what I thought you meant. That is, I thought x is the number of listeners, and y is the number of streams that those listeners consume per month. It makes sense to do it this way, since the number of streams per month is dependent upon listener behavior, and not the other way around.

        But never mind. Let’s actually use the power law with your definitions. A power law follows the form y = ax-c = a/(xc) where a and c are constants. According to you, x is the number of streams that listeners consume per month, and y is the number of listeners who consume that number of streams. Thus, the point (200, 1000) would mean that there are exactly 1000 listeners that consume exactly 200 streams per month.

        Let’s take the minimum number of streams (xmin) to be one per day, i.e. 30/month. We know from the Audiam data that the average stream per listener (mean) is roughly 700. We also know that the mean of a power law distribution can be approximated by mean = xmin * 21/(c - 1), as long as c > 1. Plugging in the numbers, this means c = 1.2201.

        Unfortunately, this does not give us a value of a. To find this, we need to integrate over all possible x, and find a value such that the sum is equal to the total number of subscribers. We need to know xmax as well. Let’s choose 20,000 streams/month; this is a listener who listens all day, every day, to songs that are about two minutes long. And we know that Spotify’s total user base is about 40 million users; let’s use that as a nice round number.

        Plugging in these numbers, this means a = 24,340,300.

        So, our “power law” equation is y = 24,340,300 / (x1.2201).

        How many listeners are there, that are “average” (700 streams/month)? If we plug x = 700 into that equation, we get about eight thousand listeners. That’s not a lot of listeners!

        How many listeners are there, that listen to your number for a “typical” user (200 streams/month)? If we plug x = 200 into that equation, we get about 38 thousand listeners. Even your “typical” user is vastly under-represented.

        How many listeners are there, that listen to twice as many streams as “typical” (400 streams/month)? If we plug x = 400 into that equation, we get about 16 thousand listeners. That’s less than half as many as the “typical” user.

        How many listeners are there, that listen to half as many streams as “typical” (100 streams/month)? If we plug x = 100 into that equation, we get about 88 thousand listeners. That’s more than double as many as the “typical” user.

        Now for the outliers. How many listeners are there, that listen to our absolute maximum number of streams (20000 streams/month)? If we plug x = 20000 into that equation, we get 137 listeners.

        How many listeners are there, that listen to our absolute minimum number of streams (30 streams/month)? If we plug x = 30 into that equation, we get over 380 thousand listeners. In other words, there are over ten times as many listeners who stream the absolute minimum number of tracks, than there are listeners who are what you claim is “typical.”

        This model is obviously broken. And, I should point out, it will be broken in exactly the same way no matter what you choose for a and c.

        There is no chance, none, zip, zero, that streaming usage will be a normal distribution (i.e. bell curve).

        Says you. You have provided no evidence (none, zip, zero) that streaming usage will not be a normal distribution. If you have evidence that this is how listeners of streaming services behave, then let’s see it.

        Now, I happen to believe that it is something close to a normal distribution, and possibly positively skewed. I’m also willing to believe that it resembles a log-normal distribution or the probability mass function of a Poisson distribution.

        But there is no chance, none, zip, zero, that streaming usage will follow a power law distribution.

        Like

      • Just for kicks, I did a log-normal distribution.

        In a log-normal curve, the mean is given by eμ + (σ^2)/2. (Sorry about using “^”; this WordPress design apparently doesn’t support nested superscript tags.) We’ll take the average from the Audiam data: 700 streams/month. The mode is given by eμ - σ^2. You claim that this is somewhere around 200 streams/month. I still doubt that number, but let’s take your word for it.

        We now need to figure out μ and σ. To do this, we can simply set the mean to seven-halves of the mode (the ratio of “average” to “typical”), and solve. First, let’s set μ to zero (which is an allowed value) and solve for σ. Plugging in the numbers, we get σ = 0.913879. This is extraordinarily positively skewed, relative to a normal distribution.

        To find the actual value for μ, we solve the equation eμ + (σ^2)/2 = 700. This gives us μ = 6.13349. (It is the same result if we solve the mode equation for 200.)

        Keep in mind that any values that come out of this will be values from 0 to 1, so we have to multiply by forty million to get the actual number of Spotify subscribers. So, here’s our PDF equation for the log-normal distribution:
        y = 40000000 *(1/(x * 0.913879 * sqrt(2π))) * e(-(ln(x) - 6.13349)^2) / 2 * (0.913879^2)).

        EDIT: I realized after I made these comments that the actual number of Spotify premium users is four million, not forty million. But the numbers come out proportionately the same, just divide by ten.

        Let’s see what we get now.

        How many listeners are there, that are “average” (700 streams/month)? If we plug x = 700 into that equation, we get 22,472 listeners. That’s not a lot of listeners – but it seems a lot more realistic than eight thousand.

        How many listeners are there, that listen to your number for a “typical” user (200 streams/month)? If we plug x = 200 into that equation, we get 57,504 listeners. That’s a lot more than under the power law distribution.

        How many listeners are there, that listen to twice as many streams as “typical” (400 streams/month)? If we plug x = 400 into that equation, we get 43,130 listeners. Less than the “typical” user – but not by half, as with the power law.

        How many listeners are there, that listen to half as many streams as “typical” (100 streams/month)? If we plug x = 100 into that equation, we get 43,130 listeners. This is exactly what we expect if we consider 200 streams/month to be the mode.

        Now for the outliers. How many listeners are there, that listen to our absolute maximum number of streams (20000 streams/month)? If we plug x = 20000 into that equation, we get zero listeners (rounding down from 0.176). This is to be expected – in the long tail, numbers like these are anomalies.

        How many listeners are there, that listen to our minimum number of streams (30 streams/month)? If we plug x = 30 into that equation, we get 6,667 listeners. Not a lot – and a lot less than the average listener. But these are people who would not listen to fewer streams, they would drop the service.

        Let’s do one more calculation with the outliers. We know that a listener must be an integer. Let’s consider the single user who streams the most on the service, and the single user who streams the least. Under the log-normal model, what are their number of streams? In other words, at what point does y = 1?

        If we plug y = 1 into that equation, we get 14,428 streams/month for the single heaviest user, and a mere 3 streams/month for the lightest.

        So, it’s not perfect. Obviously, someone who listens to a mere three streams in a month wouldn’t pay for the service. And I doubt that even a single user managed to stream over 14,000 tracks in a single month, even if they were fraudsters.

        But it’s still far, far more realistic than the power law model.

        Like

      • How funny, I did a log normal distribution too. Here’s the excel file: https://drive.google.com/file/d/0BxeQhxjGaG_9VUpGMS05Y19WZEk/view?usp=sharing
        I tried to upload it to Google spreadsheets but it crashed trying to make the graph, so I deleted the graph:
        https://docs.google.com/spreadsheets/d/1VnutQ2OppC1tklm1CuBKJ92klZcXA1dKBk-2KcQqigQ/edit#gid=29850269

        I’ve already upgraded my streaming model to include these log normal numbers.

        When I was referring to a power law earlier I was referring to the decay. Obviously a power law wouldn’t account for the ramp up, and I agree a log normal distribution is probably going to be a more accurate description of what reality looks like.

        I still believe the 80/20 rule is quite likely in effect though.

        So, it’s not perfect. Obviously, someone who listens to a mere three streams in a month wouldn’t pay for the service. And I doubt that even a single user managed to stream over 14,000 tracks in a single month, even if they were fraudsters.

        You probably didn’t see my reply below about data mining last.fm scrobbles to find some examples of usage:

        Obviously, someone who listens to a mere three streams in a month wouldn’t pay for the service
        This guy (Neil Gaiman, a famous author, who I know for a fact is a Spotify subscriber) did a whopping 17 streams in February:
        http://www.last.fm/user/neilhimself/library?from=1422748800&rangetype=1month
        He averages just 1500 streams a year:
        http://www.last.fm/user/neilhimself/library

        And I doubt that even a single user managed to stream over 14,000 tracks in a single month
        This guy was *averaging* around 18,000 tracks a month.
        http://www.last.fm/user/alkl/library?from=1262304000&rangetype=year
        All songs on the Fat Wreck Chords label, mostly NOFX. Seems legit.

        Trolling around on Last.FM I can find cases where users didn’t stream at all in a month. Maybe they were traveling, or got sick, or acquired other interests for a time. Music isn’t the only way thing people use to entertain themselves, and just because they subscribe isn’t a guarantee that they are going to use it every single month.

        I know as well that the majors are *obsessed* with getting breakage (fees collected when there are zero streams), because they don’t have to pay those out to artists.

        Like

      • You probably didn’t see my reply below about data mining last.fm scrobbles to find some examples of usage:

        I did see that. But first of all, cherry-picking outliers doesn’t prove any kind of point. (Though I guess it is possible to listen to a huge amount of songs – but I will note that this users hasn’t been active in the past two years. Was he doing shoutcasts, maybe? Who knows.)

        Second of all, there are a ton of reasons why Last.fm wouldn’t scrobble a user’s listening behavior. Last.fm requires a plug-in to your music player to do that, you have to explicitly set it up in Spotify, and it’s not well-supported in other services. Personally, I stopped using it a while ago.

        BTW, their data goes to Echo Nest (along with other services’ data), which is why I hoped that the Echo Nest data would be more usable.

        Third, I really take exception to using phrases like “worthless” when talking about light users. Someone’s worth is not determined by the royalties they generate.

        Fourth, whatever light users listen to will result in payouts. They’re not “worthless.” But under your model, light users would be the only users that actually generate any royalties. They are the only ones that are not “worthless” (using your term).

        Like

      • How funny, I did a log normal distribution too.

        Just so you know, that log-normal distribution is vastly and unrealistically skewed.

        For example, the mode (“typical” listener) would listen to less than 100 streams per month. This is less than half of what you claim as “typical” (200-300 streams), and I believe that even what you claim is “typical” is too low. This is clearly not accurate.

        How did you get the numbers for μ and σ? I suspect that you chose them to fit the “80/20” rule, which means that you are lying with statistics.

        Liked by 1 person

      • How did you get the numbers for μ and σ? I suspect that you chose them to fit the “80/20” rule, which means that you are lying with statistics.

        Actually it doesn’t fit the 80/20 rule. 20% of the users consume 64% of the streams. I thought you would be happy about that?

        I downloaded the graph from here: http://www.quantitativeskills.com/sisa/rojo/distribs.htm

        I just played with the graph until the average was 720 streams per month, it took me a while to figure out how to do that, and I needed to make a lot of changes. I tried to make the population 100k too but that crashed excel. After I got to an average of 720 I did spend 15 minutes trying to get the top of the curve to be wider and a little more to the right, but I was out of steam by then. Maybe you can help?

        Like

    • “But first of all, cherry-picking outliers doesn’t prove any kind of point.”

      Just to be clear, I didn’t spend hours looking for outliers to cherry pick. I spent five minutes scrolling through a small handful of queries. I probably looked at less than a dozen in total.
      Here’s a more recent heavy user: http://www.last.fm/user/NMH1998/library?from=1388534400&rangetype=year

      “Though I guess it is possible to listen to a huge amount of songs”

      That indeed was the point I was trying to make. I was also making the point that many listeners are not listening to any music in certain months.

      “but I will note that this users hasn’t been active in the past two years”

      Last.FM cuts you off when you reach a million streams. I find it hard to believe that this user had any purpose other than fraud.

      Second of all, there are a ton of reasons why Last.fm wouldn’t scrobble a user’s listening behavior.

      Yes, I know. It also scrobbles just about anything, and we can’t tell what the source is. It has deep limitations. But I hope you agree with me that it’s hard to come up with a legitimate reason as to why someone would be scrobbling a half million tracks a year.

      BTW, their data goes to Echo Nest

      Spotify owns Echo Nest.

      Third, I really take exception to using phrases like “worthless” when talking about light users. Someone’s worth is not determined by the royalties they generate.

      Let’s take Neil Gaiman. He streamed Zoe Keating one month, and that’s all he did.
      Now let’s take NOFX guy. He streamed NOFX and that’s all he did.

      If we hypothesize that they were both on Spotify, then Zoe got 7 cents, and NOFX got $291.

      Neil is pretty much worthless here, sorry. 7 cents doesn’t do anything meaningful for Zoe. There isn’t a whole lot you can buy for 7 cents.

      If Neil wanted to support Zoe, and I’m pretty sure he does, his “vote” basically doesn’t count. So to the extent he wants to support her, he can’t, without expending an enormous amount of effort. Even worse, Fat Wreck Chords manipulation of the system is making Neil’s pitiful gesture worth even less.

      Democracy runs on the proposition that every person gets one vote. Spotify runs on the proposition that there are no people, there’s just votes. So if you want your candidate to win, you have to vote a lot. Some people don’t have time to vote a lot. So the end result is some people matter a lot, and other people don’t matter at all.

      “But under your model, light users would be the only users that actually generate any royalties.”

      This is simply not true. Under Subscriber Share every user generates $7 of royalties. Every user is worth exactly the same, no one is worth more or less than anyone else. If you want royalties you need to get fans and capture mind share. If you have a lot of fans, and good mindshare, it won’t matter if they are light or heavy you will do great.

      Like

      • I find it hard to believe that this user had any purpose other than fraud.

        Of course he could have, and I already listed one. He could have been streaming a Shoutcast station (which means Last.fm records what he plays, even if nobody is listening). Given the fact that his streams are from 2010 – before Spotify was even available in the U.S. – I find this much more likely than “fraud.”

        Spotify owns Echo Nest.

        That doesn’t mean that Spotify streams are the only ones recorded by Echo Nest. Last.fm makes it clear that this is where their data goes, and they’re not alone. (And those services started sending data to Echo Nest long before Spotify acquired them.)

        So I guess I don’t see the point you’re making.

        Let’s take Neil Gaiman.

        Neil’s worth is not determined by how much he generates in royalties. Even from an artist-centric viewpoint, he is “worth more” if he adds your song to his playlist, or tweets about you, or recommends your band to Amanda. But from a human standpoint, his “worth” is derived from his talent as a writer (of both novels and graphic novels), or a screenwriter for Doctor Who, or as a father and a husband.

        No artist should give a rat’s ass how much he generates for them in royalties.

        If we hypothesize that they were both on Spotify, then Zoe got 7 cents, and NOFX got $291.

        Right. But let’s say you had your way. If Neil listened to Zoe once, she would get $7.00. If the NOFX guy listened to Zoe once, she would get .00039. Even though she is equally as popular with both listeners.

        For listeners who want to support artists, the undeniable lesson is that they should listen to a lot less music. Otherwise they won’t generate royalties for artists; they would be, by your standards, worthless.

        Like

      • “Given the fact that his streams are from 2010 – before Spotify was even available in the U.S.”

        Sure it could be a Shoutcast station, it could also be somebody hacking Last.fm, or it could be signals from aliens that Snowden hasn’t decrypted yet. But this all feels very silly: why are you jumping through hoops to minimize the possibility of fraud?

        I’ve shown you links to people willing to stream music 24/7 on Spotify in exchange for money.
        I’ve shown you apps (such as Eternify) people built specifically to enable fraud.
        I’ve even provided links to blog posts where fraudsters in their own words described how they did it.
        But for some reason you really want to minimize it.

        No artist should give a rat’s ass how much he generates for them in royalties.

        Then the converse must also be true: why do you give a rat’s ass what heavy listeners generate in royalties? Why are you so committed to making sure they get to spend other people’s money?

        Right. But let’s say you had your way. If Neil listened to Zoe once, she would get $7.00. If the NOFX guy listened to Zoe once, she would get .00039. Even though she is equally as popular with both listeners.

        I’ll make you a deal: I’ll answer this question, but first tell me why you think NOFX guy should get to distribute $284 of other people’s money to NOFX?

        Like

      • why are you jumping through hoops to minimize the possibility of fraud?

        Because these plays are from 2010, meaning it’s almost certainly what this guy listened to on his music software. These plays aren’t generating royalties for anyone. There’s no financial motivation for fraud.

        It is much more believable that this guy was someone who had a Shoutcast station, or something similar. Maybe it was personal, maybe not (he could be running some sort of “official station” for Fat Wreck Chords or something).

        Incidentally, most of the other “fraudsters” probably aren’t in it for the royalties, either. Whatever these guys pay people, it’s probably going to end up being more than what those people generate in royalties. It’s likely for other reasons (e.g. “pay for promotion” schemes). Also, Spotify shut down Eternify almost immediately. (And the app was not so much a “scam” as a protest about how “music streaming’s virtually worthless for artists.”)

        I don’t approve of this sort of thing either, but a “subscriber share” model won’t do much to slow these guys down.

        These things happen – and will happen no matter what – but they’re outliers. They’re not the typical above-average listener, and have a negligible effect on payouts overall. Making the discussion all about them is just a red herring.

        Then the converse must also be true: why do you give a rat’s ass what heavy listeners generate in royalties?

        Obviously, I think artists should consider how much they are paid by streaming services. My point is that individuals shouldn’t be judged solely on the amount of royalties they do or do not generate. Even if an individual generates absolutely no royalties for me, that doesn’t mean they’re “worthless.”

        Neil, for example, could never be “worthless,” because he is a great writer, he was funny in the Guild, he has what I think is a great outlook on the whole “piracy” business, and because he’s married to someone I know and like. Calling him, or people like him, “worthless” just because they don’t generate much in royalties, is insulting.

        Choose a word that’s not insulting.

        tell me why you think NOFX guy should get to distribute $284 of other people’s money to NOFX?

        Because it’s not other people’s money. It’s part of the pool of money that users pay to listen to music however they want. Phrasing it this way is an obvious appeal to emotion.

        Do I think it’s unfair that streaming services pay out according to how people use the service, rather than by how much people pay for the service?

        Nope. That’s what happens with any “all-you-can-eat” service, in some way or another.

        Do I think it’s unfair that streams from people who listen to less music are paid the same as streams from people who listen to more music?

        Nope. The heavier listeners may generate more royalties than the lighter listeners, but that’s because they listen to more music. This is a moral good; they should never be discouraged from listening to more music, and the artists who they listen to shouldn’t be paid less because of it.

        Do users think it’s unfair? I’m sure some do; I think most don’t. But if they do, then their response should be either 1) “I shouldn’t be paying as much as the other guys,” or 2) “I’d better listen to more music, to get my money’s worth.” Of course, both responses result in fewer payouts to rights holders.

        Why do I not like the “subscriber share” model?
        a) It gives more money to rights holders who are already the top earners, and takes away money from rights holders who already earn less. It makes the system even more top-heavy than it already is.
        b) It makes lighter listeners (i.e. casual listeners) disproportionately powerful in determining who gets royalties, and makes heavier listeners (i.e. music fans) almost completely powerless in this regard.
        c) Because of b), it provides a financial incentive for rights holders to make users listen to less music – meaning, less of other people’s music. It creates a zero-sum game, and encourages rights holders to lock each other out.

        I doubt that artists themselves will act on option c). But you can bet that rights holders – especially the Big Three rights holders – certainly will. Collectively, they have the power to lock out anyone who isn’t in the Big Three, and to monopolize promotion to the “casual listener” market. They’ve been doing this for years, and are already motivated to do so. They don’t need yet another reason.

        Like

      • “Because these plays are from 2010, meaning it’s almost certainly what this guy listened to on his music software.”

        Rhapsody has been around since 2000 (I was working there when it launched), and has had deals with the majors in place since 2003. Regardless there’s plenty of other more recent examples of excessive plays, and even fraudsters stating in their own words how they committed fraud.
        This guy has been 450k scrobbles spread among just 6 artists, mostly Neutral Milk Hotel. The scrobbles are 24 hours a day, 7 days a week.
        http://www.last.fm/user/NMH1998/library?rangetype=year&page=1&from=1388534400

        Look at June 2015, does this look legit to you?:
        http://www.last.fm/user/NMH1998/library?from=1433116800&rangetype=1month

        July 26th he stops suddenly. He restarts one day on August 13th, and then goes silent again on the same day. Banned? Shut down his shoutcast station? Maybe.

        He did 120,000 tracks in 2014. If he’s on Spotify he paid $120 in fees, and generated $840 in royalties.
        In the same year Neil did 900 tracks. Neil paid $120 in fees and generated $6.30 in royalties.

        You can come up with a lot of reasons why Last.FM data isn’t reliable (and I’ll agree with you, it’s not), but what you can’t say is that it’s impossible, and that people aren’t committing fraud.

        “Whatever these guys pay people, it’s probably going to end up being more than what those people generate in royalties.”

        There’s people on fiverr every day offering to stream your band for money on a cash flow positive basis.
        This guy is charging $40 for 12,000 plays. That’s worth $84 in royalties. Note he’s got a ton of satisfied customers giving 5 star reviews.
        https://www.fiverr.com/streamify/get-you-1-million-spotify-plays?context=advanced_search&context_type=rating&pos=6&funnel=7e2545ac-2b26-4ae2-9ab0-a9ddb251a23f

        “I don’t approve of this sort of thing either, but a “subscriber share” model won’t do much to slow these guys down.”

        That’s simply not true. The financial incentive for click fraud goes to zero under subscriber share. Let’s say I get an artist to pay me $40 for 12,000 plays. The artist will get $7. If I split it up among 3 subscription accounts he’ll get $21, and I’ll make a $10 profit. If I split it up among 6 subscription accounts he’ll get $42, and I’ll lose $20.
        There might be artists who are dumb enough to pay for play counts for other reasons, but the financial upside goes to zero.

        I just spoke yesterday with a well connected digital music industry person and he was an adviser for a startup that worked on the click fraud problem. He said they were shocked at the numbers they saw. Much bigger than anyone realizes. But found no one was willing to pay for their work. There’s no financial incentive for the services, it’s not their money.

        This is a real problem, and denying it exists isn’t going to solve it.

        “Because it’s not other people’s money. It’s part of the pool of money that users pay to listen to music however they want. Phrasing it this way is an obvious appeal to emotion.”

        No it’s an appeal to logic. The first thing, THE FIRST THING, they teach you about debating economics is this rule:

        People do more of something when the reward increases. When you subsidize something, you get more of it.

        When listeners give money to artists, they are “voting” for more music from that artists. When we distribute listener’s money to artists they don’t listen to we are robbing them of their vote, and we are giving extra votes to other subscribers who very well might not be listeners at all.
        I see that as an inequity that is robbing both listeners of their votes, and artists of their justly earned rewards.

        Even if an individual generates absolutely no royalties for me, that doesn’t mean they’re “worthless.”

        Sigh. I am not saying they are worthless as people, I’m saying their votes don’t count in any meaningful sense of the word, and the royalties they generate are negligible despite the fact that they have tremendous value as fans, and as individuals. If you make an effort to be fair with what I’m trying to say, I will do the same, and not parse your sentences to a ridiculous point where the plain and clear meaning is lost.

        Obviously, I think artists should consider how much they are paid by streaming services.
        No artist should give a rat’s ass how much he generates for them in royalties.

        These statements are mutually exclusive in my view.

        “Do I think it’s unfair that streaming services pay out according to how people use the service, rather than by how much people pay for the service?
        Nope. That’s what happens with any “all-you-can-eat” service, in some way or another.”

        I’m fine with customers paying the same amount regardless of usage. The problem is paying the workers based on usage. No one in an AYCE buffet restaurant, not the employees, not the owner, not the vendors, is paid on a per-bite basis. The restaurant’s revenue is contingent on getting more people in the door, not more bites of food.
        If a buffet rewarded chefs, employees and vendors only for bringing in the heaviest eaters, and paid pennies for light eaters it would go out of business.
        In an AYCE buffet the focus is on getting light eaters to show up. And heavy eaters even get kicked out periodically!
        http://www.forbes.com/sites/modeledbehavior/2012/05/23/the-economics-of-all-you-can-eat-buffets/

        Let’s go back to the spreadsheet. If an artist is equally popular among all deciles EVEN UNDER A NORMAL DISTRIBUTION they will make more money under subscriber share. They only lose under subscriber share if their fans are deeply tilted towards the top 10% of heavy users. It seems unlikely to me that a brand new emerging artist, the one that needs the money the most, and will put it to the best use, would typically develop a fan base of only people in the top 10-20% of usage. It seems much more likely they would have a spread of listeners similar to what we find in the population of users at large. So if they have 100 fans, then 10% might be in the category of 10% heaviest users, and 10% would have usage comparable to the 10% lightest users. In this case the artist would be better served by Subscriber Share. How much better served would depend on how skewed to the right the data was. This is indeed what Pedersen’s analysis showed: local artists did substantially better under subscriber share.

        “The heavier listeners may generate more royalties than the lighter listeners, but that’s because they listen to more music.”

        Let’s apply this to an AYCE buffet: Heavier eaters may generate higher pay for employees than lighter eaters, but that’s because they eat more food.
        Does that make sense? Nope.

        “Do users think it’s unfair? I’m sure some do; I think most don’t. But if they do, then their response should be either 1) “I shouldn’t be paying as much as the other guys,” or 2) “I’d better listen to more music, to get my money’s worth.” Of course, both responses result in fewer payouts to rights holders.”

        My experience is most users are completely unaware of how the royalties are distributed. They are surprised when they learn.
        1) As I’ve said many times, I would have zero problems with Pro Rata if users were also paying by the click.
        2) The value of having your vote count is a bit abstract, and I’ll be the first to admit that’s a harder concept to get across. Telling people that they have to work harder (i.e. keeping up with bots and offices) in order to make their voices heard is a tough sell. Nonetheless there are quite a few people who are upset to learn their votes don’t count. There are also many people who don’t give a shit, and consider this the musician’s problem.
        I’m asking those who do care to help the labels understand that they are running their AYCE buffets the wrong way. I’m fairly confident we’ll get there.

        “a) It gives more money to rights holders who are already the top earners, and takes away money from rights holders who already earn less. It makes the system even more top-heavy than it already is.”

        Lately I’ve been using this analogy to describe the issue:

        Imagine we have 1000 mice. And we have two gasses. Gas #1 will kill 60% of the mice. Gas #2 will kill 59% of the mice.
        Pretty easy decision, right? Gas #2 will save more mice!

        But what if I told you that 20% of the mice had a virus that would kill a lot of humans and also kill a lot of mice too. And that gas #1 would kill most, if not all, of these mice.

        Now gas #1, despite the fact that it kills more mice in the short run, saves more mice lives in the long run. It gets rid of a lot of bad mice, and let’s more of the good mice live.

        That’s what we have here. Musicians who make music good enough to attract real fans are, as we know, comparatively rare. But they are the “good mice”.
        Musicians who manipulate the system by buying clicks on fiverr, or otherwise attract a disproportionate percentage of “heavy eaters” are the “bad mice”.
        Subscriber share kills the bad mice, and saves the good mice. On paper, and in aggregate, this looks like more mice are getting killed, but it’s really best for everyone in the industry.

        “b) It makes lighter listeners (i.e. casual listeners) disproportionately powerful in determining who gets royalties, and makes heavier listeners (i.e. music fans) almost completely powerless in this regard.”

        We simply disagree here. In my view it makes everyone equally powerful: everyone gets $7, and they can divide it however they want. If heavy users want more of their money to go to a certain artist, they can stream that artist more often.

        “c) Because of b), it provides a financial incentive for rights holders to make users listen to less music – meaning, less of other people’s music. It creates a zero-sum game, and encourages rights holders to lock each other out.”

        Newsflash: music is, and has always been a highly competitive business. Labels already compete for marketshare. Artists compete for fans. I was a major label artist and you better fucking believe that every artist on every label is paying attention to how other artists are doing, noting ruefully if a peer is doing notably better than they are. Listeners vote with their ears and wallets. You can’t tell them what to do. If they hear something they like, they will listen to it. If they don’t like it, they won’t. Subscriber share doesn’t change this dynamic. It just makes sure that every listener’s vote counts.

        The vast majority of the indie artists I have spoken to or heard from (and I speak to indie artists every day, because that’s my job) see it the same way I do. Zoe Keating sees it my way. Rogue Wave sees it the way I day. Ted Leo sees it the way I do. Pavement sees it the way I do. Amanda Palmer sees it the way I do (if you know her, ask her yourself). Countless other artists see it the way I do, and have tweeted, shared, written, called and emailed to voice their agreement with my post. Furthermore the vast majority of listeners I have spoken with see it my way. Numerous “smart” people in the industry, people deeply respected in the industry like Mark Mulligan and David Lefsetz see it the way I do. The only people who have studied this issue using real data, and spending a long time looking at this data (Arnt Maaso, Rex Pedersen, and their respective teams) have publicly stated they see it the way I do. People at the services themselves have told me privately (and off the record) they support this idea. People at labels (again, off the record) have told me they support the idea. Over 150,000 people have read my blog post, but I can measure the objectors (of which you are clearly one) in the dozens.

        There is data that will definitively establish one way is right and the other is wrong. I’m here in good faith and if it’s revealed I am in error, then I will issue a mea culpa, and let it go. I don’t have any ego invested in this, and I don’t have a financial stake in the outcome. However I have not seen that data yet. And the arguments you (and the small handful of others who disagree) have presented I find unpersuasive.

        At this point I think we’ve talked it out as far as we can. I appreciate you humoring me with the discussion. If you can ever find a decent log-normal graph on excel, or if you encounter some data that you think I might be interested in, by all means let me know.

        Like

  8. I’ve been poking around Last.FM scrobbles. I was thinking they might be a way to get an idea of how much music a typical person listens to, although it’s not ideal as I would imagine people who scrobble are probably more likely to be serious music fans.

    However poking around has turned up some pretty interesting things.

    Look at this guy:
    http://www.last.fm/user/alkl/library?from=1262304000&rangetype=year
    He was going around 18,000 tracks a month. All songs on the Fat Wreck Chords label, mostly NOFX. Seems legit.

    This guy has streamed over 1M streams (we don’t know how many he’s done since reaching a million because Last.FM cuts you off when you reach a million):
    http://www.last.fm/user/RBD007/library
    All of his streams were split between two artists: RBD and Anahi. Ever heard of them? Me either….
    http://www.last.fm/user/RBD007

    On the other hand look at this girl: http://www.last.fm/user/padarnalat/library?from=1420070400&rangetype=year
    She’s an indie rocker (Sloan, Guided By Voices, Yo La Tengo, Big Star) with diverse taste but her best month this year was just 350 tracks.

    Another light user with diverse taste:
    http://www.last.fm/user/LAST.HQ/library?from=1420070400&rangetype=year

    For the past 5 years this user has averaged around 1500 tracks *a year*. That’s 125 streams a month. In a pro rata world this user doesn’t matter.
    http://www.last.fm/user/neilhimself/library
    In real life he matters a lot. He’s Neil Gaiman, a pretty amazing author, and married to the famous musician Amanda Palmer.

    Like

  9. As an addendum, Spotify now has over 30 million subscribers. Given the fact that Laguana based his figures on (at the time) 4.7 million subscribers, and that he considered only 15,000 participants significant enough to piss off major labels, I’m genuinely curious if he’s still willing to cling to his ridiculous numbers. And, if not, whether he’s willing to admit he was completely wrong.

    I suspect not, but who knows, I could be pleasantly surprised.

    Like

    • My “ridiculous” hypotheses did indeed attract the attention of every streaming service, major label, and the major independents as well. We have debated the topic on stages in front of audiences in a number of locales, and privately either one on one, or in invitation-only industry mailing lists. Several of these people are now friends, people I look forward to seeing. The post did precisely what it was supposed to: attract attention to the idea. It’s been wildly successful on that front, with hundreds of thousands of views. It’s now on the curricula in several colleges, and I’ve been an invited guest lecturer on a couple of occasions.
      It’s still being debated. The primary challenge is that while it doesn’t hurt the majors or the services, it doesn’t boost their bottom line in the short run either. And right now everyone’s focus is on growth, which I think is entirely fair.
      Sooner or later though it will eventually become an economic necessity to revisit the concept that “every stream is equal/some users are worthless” and replace it with “every user is equal/some streams are worthless”. To some extent this is already happening with the rise of exclusive deals between services and artists, but a more algorithmic approach will deliver a lot better ROI.
      Artists with real fans, as opposed to bots & clickfarms, are already catching on that releasing an album on a streaming service right away leaves a tremendous amount of money on the table: an excited engaged click is not worth any more than a passive/ignored/bot/fraud click, and every one of these bad clicks decreases the value of the excited/engaged click. That excitement and engagement is far better monetized through CD’s and vinyl where all of the fan’s money goes to the artist that created the excitement. Then after a while, when things have died down, and the real fans have already bought the record, the artist can quietly release the album to streaming for passive background listening by casual fans.
      These artists would be far less likely to do this in a subscriber share world. And sooner or later a service will emerge that understands this and capitalizes on this opportunity.
      In the meantime it’s been both interesting and rewarding sticking up for emerging artists. I know you don’t agree but whatever, you just enjoy hating on things. Even your blog name is not trying to advance any idea of your own, but is instead just a way to try and tear someone else’s ideas down a notch. Amazing how much energy you’ve put into that. And a bit sad. Maybe you could find some way to make an actual contribution instead of just trying to tear other people down? Imagine the things you could do! A lot more interesting than being the streaming music equivalent of the comic book guy.

      Like

      • Hello again, Sharky.

        My “ridiculous” hypotheses

        I said the numbers were ridiculous. You explicitly based the success of your campaign on the premise that it would hurt the major labels’ bottom lines. If this didn’t happen, then it was not a success as you defined it. And, from what you just said, it appears it was not:

        The primary challenge is that while it doesn’t hurt the majors or the services, it doesn’t boost their bottom line in the short run either.

        On the other hand, I also said that people should participate nonetheless, because it may help the indie artists that the participants listened to. You have not mentioned whether this is the case. (Did those indie artists make more money? Were their chart positions on Spotify bumped up?) I suspect you don’t have this data, but if you are selling this as beneficial to indie artists, you should find out. And if you can’t, then admit that you don’t know whether it helped or not.

        The post did precisely what it was supposed to: attract attention to the idea. It’s been wildly successful on that front, with hundreds of thousands of views. It’s now on the curricula in several colleges, and I’ve been an invited guest lecturer on a couple of occasions.

        That’s great for you. I’m still not sure it’s good for indie artists (or artists in general). Attracting attention to the idea is only a good thing if it’s a good idea, and I’m still waiting for evidence that it is.

        The one thing that should have come out of all of this, is that more people would be interested in releasing and analyzing the actual data from Spotify and other streaming services, so that the idea can be tested. Has that happened as a result of your campaign? If so, then you definitely should be congratulated for that… and of course, I’d love to see the data.

        Artists with real fans, as opposed to bots & clickfarms, are already catching on that releasing an album on a streaming service right away leaves a tremendous amount of money on the table: an excited engaged click is not worth any more than a passive/ignored/bot/fraud click, and every one of these bad clicks decreases the value of the excited/engaged click. That excitement and engagement is far better monetized through CD’s and vinyl where all of the fan’s money goes to the artist that created the excitement. Then after a while, when things have died down, and the real fans have already bought the record, the artist can quietly release the album to streaming for passive background listening by casual fans.

        I see you’re still going ahead with this characterization, and I still wish you wouldn’t do it.

        Most users who stream more music than average are not “bots” or “clickfarms.” A “passive” click is not a “bad” click, and is probably done as much by “real fans” as by “casual fans.” Characterizing the legitimacy of music fans according to how much they can be “better monetized” just seems wrong-headed to me. And if someone buys a physical copy, that doesn’t make them any more “engaged” or “excited” than another person who waited until the music is available on their streaming service.

        Obviously, artists need to make money, but let’s not delve into false dichotomies or “no true Scotsman” fallacies because of it.

        sooner or later a service will emerge that understands this and capitalizes on this opportunity.

        Perhaps, but given how much trouble the streaming services have right now, I’d say that’s the least of their concerns.

        In the meantime it’s been both interesting and rewarding sticking up for emerging artists. I know you don’t agree but whatever, you just enjoy hating on things.

        That’s just it: I don’t believe you are sticking up for emerging (or indie) artists. That’s certainly your stated intent, and I believe you’re genuine, but I think your plan will actually hurt them. So, in my view, I’m the one sticking up for emerging or indie artists.

        Obviously you disagree, but don’t say I’m just hating on things. I “hate” on things that I think are bad ideas. And in my view, calling out bad ideas is a service that ultimately helps everyone. (FWIW, your ideas are not even remotely as bad as the others I blog about.)

        Yeah, I probably do spend a little too much time being reactive rather than proactive. But I don’t like people who promote bad ideas in the name of “helping” artists, and I feel the need to speak out against them. Especially if these bad ideas actually hurt artists, and most especially if they result in bad policies that hurt everyone. As both an (extremely underground) artist and a member of “everyone,” I consider this to be making a contribution.

        A deliberately extreme example: There are a ton of websites out there that spend an awful lot of time and energy “tearing down” the anti-vaccination movement. Do you think they’re not making an actual contribution? Do you think they should be doing something else?

        Obviously the things I talk about here aren’t on the same level of bad as the anti-vaccination movement. (Especially not your ideas.) But I see myself in the same general category: speaking out against things I think are harmful.

        The fact that I enjoy it is just icing on the cake.

        BTW, once I get my studio set up, I’ll probably spend even less time writing blog posts, and more time creating music again. That probably doesn’t count as “making a contribution,” but at least you’ll have fewer things to be annoyed about. (Unless you actually listen to my music, of course.)

        Anyway, that’s all I have to say. Enjoy the long weekend.

        Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s